Alma Mater Studiorum

Università di Bologna

SCUOLA DI INGEGNERIA E ARCHITETTURA

Corso di laurea in ingegneria informatica

magistrale

Mission scheduler for

a rail guided vehicle system

Relatore: Presentata da:

Chiar.mo prof. Roberto Giuliani
Silvano Martello

Correlatore:
Prof. Michele Monaci
Ing. Claudio Gambetti

Sessione II
Anno Accademico 2018/2019
 To my family

1
Abstract

A transport system with AGVs (automatic guided vehicles) is a fully auto-

matic system that provides logistics services in industrial environments such
as warehouses and production plants. These systems have reached such a
degree of maturity as to allow, in their daily use, the application of heuristic
algorithms for the optimisation of the various operations they perform. For
instance find the shortest paths between working stations and storage area,
assign movements and strategic positions for idle vehicles, operate efficient
and long-life battery management and more.
A relevant interesting algorithm, presented and developed in this thesis, con-
cerns the sorting of products in the shipping phase, which affects the schedul-
ing tasks assigned to the autonomous vehicles. The scheduler has the aims of
determining which operations have more strict constraints and more priority
over others.
Studies and practice have shown that the adoption of a valid scheduler im-
plies considerable improvements in the system performance, consequently it
is advisable to dedicate time and effort to the research for the right one.
The following algorithms obtained a successful outcome and they have been
implemented for the production of a modern automated warehouse located
in the city of Cesena, Italy. The paper is divided into four chapters, with a
further one dedicated to conclusions.

2
Introduction

Nowadays logistics, concerning the industrial business, constitute a funda-

mental component of the company operating system, together with produc-
tion, marketing and finance.
In recent years, the common awareness with regard to logistical issues has
grown significantly and the manufacturers are more and more aware of the
need to improve certain aspects, minimising the stock-taking and maximising
the products rotation. In this context, we find issues related to the handling
flow, which, in order to keep pace with the needs of the market, requires
state-of-the-art structures and machines.
Automation meets these demands by offering products and techniques that
can increase corporate value, reduce costs and increase revenues. An example
of this is found in AGVs systems, capable of handling hundreds of products
per hour in relation to the plant layout and the number of vehicles, reducing
maintenance costs and making the warehouse continuously operational.
This innovative internal transport system has been experiencing considerable
interest in recent years thanks to its flexibility and efficiency.
The first part of this thesis is aimed to clarify the present literature concern-
ing logistics and automated warehouses along with their respective opera-
tions, such as acceptance controls, storage, order preparation and shipment.
After providing a theoretical overview, the design and implementation of an
autonomous guided vehicles system is examined, considering its components
and formulating algorithms for their efficient control and administration.
The paper then concludes with the formulation of a scheduler, as a support
to decision-making process in the optimal management of limited resources:
the vehicles; their usage and assignment of certain activities. The approach
adopted is based on breaking down the problem into sub-problems: the first
part involves the implementation of an algorithm to define the priority of the
activities in relation to defined criteria coming from external systems; the

3
second keeps track of the current state of the system through a scheduling
matrix and decides which operation to perform when one or more resources
become idle; and, in conclusion, the third part has the task of assigning ac-
tivities to the vehicle fleet.
The heuristics of this thesis constitute an attempt to propose feasible so-
lutions that have good quality, even if not optimal, calculated on real ap-
plications and data; and through models that avoid certain simplifications
occurring in theoretical studies and academic researches.
The final part of the paper concludes with the model results as a whole and
as each individual algorithms, in different instances of the problem, from
situations of low product movement to peaks of high movement. Which are
then analysed and discussed.
The project was developed with the support of the automation business unit
from Onit group and was able to test its application in a real business sce-
nario thanks to the construction of the new Orogel plant based in Cesena.
Orogel is an agricultural cooperative company, leader in the production of
frozen foods, oriented towards innovation and in continuous development.
The year 2020 will see the launch of a new automated refrigeration plant,
the largest product storage warehouse in Europe. The warehouse is managed
in a completely automatic way, it has a surface area of 7,800 square metres,
approximately 230 thousand cubic metres of volume, and it is made to store
35 thousand tons of frozen product at an operating temperature of minus 25
degrees.

4
Contents

1 Logistics 7
1.1 Logistic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2 Logistics
integration and automation . . . . . . . . . . . . . . . . . . . 8
1.3 Supply Chain Management . . . . . . . . . . . . . . . . . . . . 9
1.4 Industrial warehouses . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 The operational
cycle of warehouses . . . . . . . . . . . . . . . . . . . . . . . . 13
1.6 Internal movement . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7 Logistic
performance measures . . . . . . . . . . . . . . . . . . . . . . 16

2 RGV system:
design and control 19
2.1 AGV vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2 RGV - Rail Guided Vehicle system . . . . . . . . . . . . . . . 22
2.3 Guide-path design:
single loop type . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.1 Compute distance to target position algorithm . . . . . 28
2.3.2 Get section from position algorithm . . . . . . . . . . . 29
2.4 Communication
between entities . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5 Material handling . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5.1 Loading and unloading preconditions . . . . . . . . . . 33
2.5.2 Handshake . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.5.3 Finite state machine . . . . . . . . . . . . . . . . . . . 37
2.6 Idle vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.6.1 Assign strategic missions algorithm . . . . . . . . . . . 43

5
CONTENTS CONTENTS

2.7 Re-positioning management . . . . . . . . . . . . . . . . . . . 44

2.7.1 Re-positioning request algorithm . . . . . . . . . . . . 45
2.8 Position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.8.1 Barcode value validation algorithm . . . . . . . . . . . 48
2.9 Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.9.1 Speed management algorithm . . . . . . . . . . . . . . 51
2.9.2 Speed management example . . . . . . . . . . . . . . . 53

3 Mission scheduler 56
3.1 The scheduler . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2 Scheduling terminology . . . . . . . . . . . . . . . . . . . . . . 58
3.3 Online scheduling . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.4 Types of algorithm -
the dispatching rules . . . . . . . . . . . . . . . . . . . . . . . 62
3.5 Mission scheduler . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.5.1 Mission data type . . . . . . . . . . . . . . . . . . . . . 66
3.5.2 Sorting products algorithm . . . . . . . . . . . . . . . . 68
3.5.3 Mission worker . . . . . . . . . . . . . . . . . . . . . . 72
3.5.4 Assignment algorithm . . . . . . . . . . . . . . . . . . 74
3.5.5 Mission status updater . . . . . . . . . . . . . . . . . . 77

4 Results 78
4.1 Sorting products algorithm . . . . . . . . . . . . . . . . . . . . 79
4.2 Mission worker . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.3 Assignment algorithm . . . . . . . . . . . . . . . . . . . . . . 82
4.4 Attachments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

6
Chapter 1

Logistics

The need to orient company production policies to the market has made
the traditional logistics and production-planning model inadequate and ob-
solete.
Today the role of logistics is to bring production closer to sales, creating
the assumption for a perfect harmony between the two operations, to en-
sure delivery of the products on time and in the quantities requested and
minimising stock, transport and information management costs, regarding
production flows.
Companies work to develop logistics projects not only with the aim of re-
ducing costs, but to increase revenues and, in general, their business: it
consequently becomes a competitive point and can represent a strength in
the market.

7
1.1. Logistic 1. Logistics

1.1 Logistic
Adopting a widely shared notion, business logistics is defined as the process
of planning, implementing and controlling the flow and storage of materials
and, additionally, of the associated information, from the point of origin to
the point of consumption, in order to respond to customer requests in the
most efficient and effective way for the company.
More precisely, the logistic process aims to harmonise flow - both of physical
products and information - through the various functions and company pro-
cesses, in order to create greater value for the company and for the customer.
Logistics cannot overlook the dynamics of market demand which, at different
levels, guide and dictate times and procedures of company operations.
Following this line of reasoning, the planning is distinguished according to
the initiators of the business logistic process, in which the strategy defined as
”pull” is the driving force, where logistic processes are activated by impulse
of the request, in response to a customer order.
Although this type of planning is common in engineered production pro-
cesses, more and more companies are adopting these productive and logistic
aspects. The result is a customer-oriented and guided business logistics, able
to respond reactively to the volatility of demand, classic of current markets.
The environment of accelerated rhythms and greater scale, in which logistics
operations must act, have led, and are still leading, to a change in the busi-
ness logistic planning model. To maximise the value in a logistics system,
there are a large number of planning decisions to consider: from the choice of
the storage system, the elements to be taken in succession to respond to the
orders received from customers and so on, until high-level business decisions
to invest in the construction of a new plant.
There is much literature and many software support tools that focus on ev-
ery single area of logistics systems. From this premise, we can understand
the importance of the supply chain in today’s world and how companies are
interested in possible improvements in any of its many aspects.

1.2 Logistics
integration and automation
The natural evolution of the logistic system leads to the development of in-
tegrated logistics, with the aim of seeking the best coordination between the

8
1.3. Supply Chain Management 1. Logistics

area that are part of it and the information system that guide its progress.
Over time, we have witnessed an ever-stronger integration of logistics flows,
via automation applied to both physical aspects and information aspects. In
fact, robotics and automation are now key success factors and are revolution-
ising the world of industry. Applied to autonomous robots and automatic
storage systems, to detection and tracking technologies or even advanced
software for the supply chain, they represent a turning point that allows an
increasingly rapid, safe and error-free distribution, shorter time to market,
with estimated efficiency increases of 20-30% and, ultimately, lower costs for
businesses and families. It should be noted that 10-15% of the cost of the
product is influenced by the logistic aspect, therefore reducing these values
ensure an increased strength in the market.
Today, logistics automation projects have reached an average annual increase
of 7% of total turnover and a long growth outlook sees this figure being close
to or above 10%. [Source ANIE Federation]
The integration of all operational activities is the main tool to achieve the
objectives set, as it reduces inefficiencies and downtime. Innovation and busi-
ness acceleration look to a rapidly changing industry. Notwithstanding the
importance of automation in the modern industrial vision, it is necessary to
maintain a critical approach: even for automatic devices, the guiding spirit
must be the technical-economic justification of each solution, thus avoiding
extreme implementations, driven by an unreasonable desire for automation.

1.3 Supply Chain Management

The term supply chain management (SCM) was introduced in 1982 by two
consultants, Oliver and Weber, to describe the connections between logistics
and other business functions. Although the principles remain the same, the
definition undergoes various transformations in relation to the area of inter-
est. Among the various possibilities, a new interpretation is preferred in view
of integration between different processes and entities.
Supply chain management is a complex system that optimises the coordina-
tion between the processes of suppliers, producers and distributors - in other
words the totality of the organisations that are involved - through upstream
and downstream connections, in the various developments that generate value
for the end customer, in terms of product or service.
SCM also includes the logistic system that starts from the raw materials,

9
1.3. Supply Chain Management 1. Logistics

transforms them into products and ends with their distribution to the fi-
nal consumers. Below is a possible outline given by the logistic chain of a
generic company, where the raw materials are delivered by the suppliers to
the manufacturing plants, which then produce components or semi-finished
parts. The finished products are then assembled in different plants, called
assembly plants. Distribution is also subdivided into several steps: two cen-
tral distribution centres (CDC) supply multiple regional distribution centres
(RDC).
The chain is created according to the type of product and request, removing
or adding plants and connections between the various phases. Moreover, the
latter can be more or less complex, composed by a single and simple line of
transport or by a more articulated network that involves different systems
and companies, such as automotive and rail.

Figure 1.1: A supply chain example

The two main flows have opposite directions: the physical one, from the
supplier to the customer and the informative one in the opposite direction.
The products follow the sequence of the supply chain, except for those that
are obsolete, damaged or do not work, while, on the contrary, information
follows the reverse path, from consumers to suppliers.
In the pull system, sellers take charge of customer requests and transmit
them to the producer who reformulates them as orders for suppliers. Simi-
larly, in pull type systems, historic sales are used to make future economic

10
1.4. Industrial warehouses 1. Logistics

forecasts on product demand and the related demand for materials. Almost
all the authors of the literature, dealing with supply chain management in
terms of process integration, agree that it:

1. allows for the opening up of a consideration of process integration, from

an intra-organizational perspective to an inter-organizational one;

2. potentially involves a very large number of autonomous organisations.

This link is inter-company in nature and can take place both in terms of
information sharing and in terms of coordinating operational processes;

3. the purpose is to supply a product/service of high quality for the con-

sumer with an appropriate use of resources and to build competitive
advantages for all the players in the chain. In keeping with such an
approach, the concept of competition passes from being that of busi-
ness against business, to one that searches for forms of collaborative
competition that increase the value for all the sides involved in the
chain.

Although some authors use the term SCM as a synonym for integrated logis-
tics, there are significant differences between the two approaches. Logistics,
unlike SCM, deals with the operations related to the movement and the flow
about goods and information. It therefore represents a function of vital im-
portance for supporting SCM, as it is directly responsible for integrating the
needs of the market with the production process.

1.4 Industrial warehouses

The warehouse must be perfectly integrated in the logistic cycle of the com-
pany. Latterly the evolution of the market and the high labour cost have
extremely complicated the role of warehouses and many companies, in order
to contain costs, have shifted production to emerging countries, making pro-
duction scheduling - and therefore the correct determination of warehouse
storage capacity - particularly complex.
The warehouse represents a key point for the supply chain, as it supplies
the main logistic service: making the right product available, in the correct
location, at the right time and at minimum cost.
The distribution system consists of two distinct structures:

11
1.4. Industrial warehouses 1. Logistics

1. the logistics channel represents the distribution network for the group-
ing, selection, sorting and transport of goods. Of these functions, the
first three are carried out within the warehouse;

2. the commercial channel is made up of the structures suitable for selling

the product.

The warehouse is important both from a temporal point of view, as it enables

economies linked to supplies and to optimise production, and from a commer-
cial perspective, making it possible to consolidate loads, with a consequent
reduction in transport costs. Warehouses are mainly divided according to
the position in which they are located: production warehouses handle large
quantities of materials and are subdivided into warehouses of raw materials,
inter-operational, finished products and combinations of them; distribution
warehouses, on the other hand, contain a wide variety of goods and are sub-
divided into central or peripheral, depending on the size of their catchment
area.
Central warehouses allow for the storage of large quantities of materials,
with the aim of maximising efficiency of reception, storage, extraction and
shipping. Regional warehouses are centres located near the point of final
distribution. They contain large quantities of goods received and stored; on
the contrary, shipments are reduced and heterogeneous.
The nature of a warehouse deeply influences the processes that manage it.
In this regard, to increase efficiency, the management of flows is totally del-
egated to automated systems. The scheduling of storage and retrieval mis-
sions is usually managed by electronic computers, equipped with dedicated
software. Minimising routes, maximising volumes within the storage loca-
tions, and minimising internal movements are just some of the logistical
problems that must be faced in order to boast an optimised and flexible
system. Many of these problems can be traced back to classic examples
of Operational Research, such as the Vehicle Routing Problem for trans-
portation or the Knapsack for volume management, although structural or
organisational constraints often make the study of the optimal solution very
complicated. In these cases, a heuristic approach enables the achievement of
the right compromise between calculation times and quality of the solution.

12
1.5. The operational
cycle of warehouses 1. Logistics

1.5 The operational

cycle of warehouses
The operational cycle of a warehouse can be subdivided into six phases, of
which the first, third and sixth are of primary importance:
1. arrival or reception;

2. acceptance testing, performed in the raw material warehouses after the

arrival of the goods;

3. storage;

4. preparation, possibly carried out after storage to better adapt the ma-
terials to production or transport requirements;

5. awaiting shipment, necessary in some finished product warehouses;

6. shipping.
The arrival area must be sized in anticipation of the acceptance of potentially
notable quantitative peaks of goods, due to a discontinuous entry flow; this
occurs especially in the case of goods coming from external suppliers, in the
presence of a strongly seasonal trend, or when materials coming from internal
departments or work-centres with non-regular production must be brought-
in to the warehouse.
The storage phase includes the period in which the product is released into
the warehouse whilst waiting to be sold. Its area represents the largest area
as it is prepared for storing unsold goods. It is characterised by a structure
specially equipped for proper storage and preservation.
The company is not willing to maintain a high number of excess stocks,
because it would mean an immobilisation of capital in unused resources.
Raw materials, semi-finished products and finished products should remain
immobilised in the warehouse as short a time as possible pending their use
in the transformation or sales processes. It is therefore necessary to plan
the requirements of materials and sub-components of the finished products
so that they are always where they are needed, when necessary and in the
quantities requested.
Shipment is the last stage of the cycle and includes the exit of the product
from the warehouse and its subsequent transport. It represents the smallest

13
1.6. Internal movement 1. Logistics

area of the warehouse due to the effect of order planning within the company,
which has greater room for manoeuvre in relation to the needs of the area in
question.
In the cost of the warehouse, receipt, shipment and storage account for 50%,
in equal parts.
Today in large complexes, shipping is completely automated, robotic vehicles
move on their own on physical or magnetic tracks, pick up products, load
them and position themselves for exit according to how the customer’s orders
are scheduled. The correct performance of tasks is facilitated by the use
of information support and the application of IT procedures for inventory
management.

1.6 Internal movement

Internal movement is subdivided into the picking phase, the storage phase
and the displacement between internal positions within the warehouse. Dur-
ing the picking phase it is required to know the real location of the item to
carry out the picking operation, while, in the stocking phase, the product
position is initially logically assigned and later the actual allocation is car-
ried out. Generally, the goods entering the warehouse are more homogeneous
than the outgoing ones. Indeed, orders from end customers often involve a
set of products that are more heterogeneous than those coming from suppli-
ers or from production chains.
Picking is more costly than storing, in terms of locations to be reached and
therefore of paths to be taken since it requires access to different classes of
products.
Each sales and shipping order results in the following set of operations:

1. the formation of batches, or the set of withdrawal missions;

2. the scheduling of the machines;

3. picking;

4. loading onto vehicles.

In the first point the orders are formed by lots, called batches. Optimal batch
size is estimated by balancing the picking times with the subsequent sort-
ing. The computation can be done using more or less complex algorithms,

14
1.6. Internal movement 1. Logistics

depending on the system requirements and the specific customer’s requests.

The second stage is carried out by electronic calculators which, through mod-
els and algorithms, have the task of identifying the best solution to promptly
release the products with higher priority, minimise the displacements and
guarantee an admissible batch ordering within the same shipment.
The third stage involves the physical picking of the products in relation to
each batch. While the previous phase has the requirement of ensuring fea-
sibility in the organisation of the batches, this phase has the true task of
sorting the items in order to allow a simple and correct picking process once
the final customer is reached.
Finally, the loading onto the vehicles takes place by packing the products in
pallets or containers, ordered according to the last come first served policy
and aggregating consecutive orders. The intention, to minimise costs, is to
reduce the number of bins needed to contain the objects: this translates into
problems of a geometric nature, with more or less constraints, depending on
whether the objects are stackable, and with various degrees of complexity,
based on the number of dimensions to be taken into account.
The third point, overall, is considered the most expensive, requiring about
70% of the total time. The picking phase is dealt with greater depth in this
thesis, thanks to the introduction of interesting and innovative flexible trans-
port systems that use autonomous guided trolleys, called automated guided
vehicles AGV.
The AGV machine, equipped with a calculator, moves along extremely com-
plex and flexible paths defined by the guide system, automatically exchang-
ing materials, pallets or other items, with the machining centres. The master
system provides real-time detection of fleet positions, management of trans-
port requests, according to certain priority rules, and scheduling the mission
orders for individual trolley.
The introduction of these devices increases efficiency, reduces operating costs
and optimises picking and stocking operations. In this regard, the following
chapter deals specifically with the design and control of AGV vehicles, intro-
ducing algorithms and models for their correct implementation.

15
1.7. Logistic
performance measures 1. Logistics

1.7 Logistic
performance measures
Logistics, warehouse management, internal handling, order management and
customer service represent processes that, if optimised and coordinated with
the actions of other agents, establish or at least influence some performances
that are critical for the success of the company.
In recent years, much of the management studies has focused on identifying
a set of metrics that can be used in logistics. These indicators integrate
economic-financial measures and technical-operational measures. Some are
general in nature and can be found in all processes, such as performance in
terms of cost, time, quality and value; others are typical of specific processes,
such as productivity, versatility and flexibility in the logistics chain.
Key Performance Indicators KPIs, are a number of qualitative and quan-
titative indicators that assess the business results achieved, with reference
to fundamental aspects of business management such as the level of service
guaranteed to the customer, the level of stock in the warehouse and the level
of costs for transport. The better the choice and measurement of the KPI is,
the better the control on the progress and the regulation of the objectives can
be. The main indicators aim to evaluate the efficiency of storage, studying
the surface saturation of the warehouse, and the materials handling, measur-
ing speed and accuracy of operations.
The main factor to be analysed within the warehouse area concerns han-
dling, i.e. the operation by which a material is stored in a location or picked
from the same, considering time and position domains. The related KPIs are
called lead-time and potential handling.
The lead-time is measured in terms of response. The lower the wait, the
greater the competitiveness. The lead-time or reaction time has a particular
importance in the logistics sector. Focal point of outgoing logistics, the lead-
time makes response times flexible between an order validation, its shipping
and delivery. The competitiveness of a company lies not only in its ability
to satisfy customer needs, but also in being able to reduce the time between
the product purchase and its delivery.
The formula is very simple, it represents the time difference between the end
and the beginning of the operation. Inside the warehouses, it is mainly used
to measure the timing between the product picking request and the conclu-
sion of the operation. This indicator can have very high values in relation

16
1.7. Logistic
performance measures 1. Logistics

to the number of operations and movements that the warehouse is currently

managing. As far as possible it should be weighted with regard to the daily
flow.
The potential handling is the dynamic capacity of the warehouse, expressed
in quantity, as the maximum number of incoming, outgoing and in-transit
units. This parameter takes into consideration the quantities and the vehicles
used. The actual potential handling is given by the product of the potential
of each individual vehicle, whether automatic or operator-guided, and the
number of means of transport.
3600
P H vehicle = U F · [operation/hour] (1.1)
T op
P H tot = P H vehicle · N vehicle [operation/hour] (1.2)
Where:
PHvehicle is the potential handling index for each vehicle while PHtot is the
total potential handling, UF is the utilization factor for each vehicle and in
conclusion Top is the time spent for usual operations.
In an automated warehouse, once designed and built, both the incoming
potential and the handling potential are fixed and defined.
A further point is the financial aspect, some financial indicators are essential
to control the operational costs related to the warehouse and appreciate
the speed of circulation, including the inventory turnover. The inventory
turnover index is the parameter used to assess the speed of regeneration of
the goods, indicating the number of times a complete renewal of stock takes
place in a given period of time. It is given by the formula:
Pt
Ui
IT = i=1 [product sold/time] (1.3)
Gm
Where:
IT is the inventory turnover index, Ui is the total consumption in the period
and Gm is the average stock in the period. Assuming that, during a defined
period, 1,000 pieces were consumed against an average stock of 250, the in-
ventory turnover index will be equal to 4, which means that the 250 units
have left, for sale or use, and then reintegrated 4 times. The index shows
that from the logical viewpoint the warehouse has been regenerated 4 times
during the year, or that the materials have remained in storage on average
91 days (365/4 days).

17
1.7. Logistic
performance measures 1. Logistics

In general, a high turnover rate translates into less capital invested in in-
ventories and lower inventory management costs, while a low rotation index
leads to the characteristic drawbacks of over-large stocks such market dis-
counts and internal inventory obsolescence. It is intuitive that the rotation
index should tend to be high.
The above listed are just some of the most common indicators used in logis-
tics. From the warehouse point of view, they become fundamental factors to
put right the flows that can slow down the entire logistic process. KPIs have
the aim of making a significant contribution to improving the process itself
in terms of value creation, in line with the implementation of short, medium-
and long-term corporate strategies.

18
Chapter 2

RGV system:
design and control

The most flexible and innovative automatic transport system inside the ware-
house is the one that uses automated guided vehicles AGV along a complex
set of routes. The exchange of material with working stations and assembly
centre takes place automatically and under the control of the central calcu-
lator. The plant is completely monitored by the supervision system, which
administers in real time every operational situations, manages the fleet and
automatically process data relating to production statistics. In this way, it
is possible to reduce human involvement to minimal responsibilities and op-
timise production flows, creating a safe and structured working environment
in a simple and linear way.
Indicatively in 2012, Amazon, the largest internet company in the world,
purchased 15,000 AGVs for its warehouses, with the aim of reducing lead
times and increasing the level of customer satisfaction.
Currently the use of AGV vehicles mainly includes the automotive, elec-
tronics and food sectors, but in the coming years an interesting increase for
small/medium companies is also predicted.

19
 2. RGV system:
2.1. AGV vehicles design and control

2.1 AGV vehicles

Barrett Electronics built the first AGV vehicle in 1950. In the 60s and 70s
the on-board controllers were first transistorised then replaced by integrated
circuits, thanks to this, in the following decades there was a strong increase
in deployments.
Thanks to their flexibility and efficiency, nowadays AGV systems are used
not only in classic production lines, but also in automated warehouses where
high quantities of goods are constantly in motion. Companies that produce
consumer goods, such as in the food and beverage field, have invested and
are still investing in automated logistics.
The possibility of covering 24 hours of work using a single AGV greatly in-
creases the market in countries where labour cost is high and where the return
on investment is rapid. Indeed, the purchase of an AGV system implies a
significant initial capital cost, but nevertheless it guarantees high economic
potential due to low maintenance costs, compared with conventional vehicles,
and to the possibility to operate without interruptions with reduced human
assistance.
In Europe, the countries that are most increasing the development of AGV
systems are Italy, Germany, the United Kingdom and the Scandinavian coun-
tries.
The online orders received by customer are managed through the company
system that prepares the regular flow of operations, including the internal
movement of the goods and the subsequent shipment, to achieve the sale of
the product. It assigns the designated products to the warehouse locations
and defines the route to reach them. The AGV vehicles will follow this route
so that the order is completed correctly and with the priorities needed.
The movement is defined by algorithms that are based on objective criteria
such as minimising travel time or waiting time caused by traffic congestion.
Although the layout is static, choosing the best possible route is a dynamic
decision and depends on current conditions and system forecasts.
The central calculator, in this context, has the task of monitoring the position
of the vehicles, controlling flows and guaranteeing productive maintenance of
the battery. Inappropriate direction policies can lead to congestion problems,
deadlocks and collisions.
The sequence of products and orders, on the other hand, depends on com-
plex indicators such as priority, product type, quantity of goods involved,
waiting time, etc. The need to carry out multiple orders simultaneously, due

20
 2. RGV system:
2.1. AGV vehicles design and control

to simultaneous sales flows, involves a high level of organisation and manage-

ment. This task is performed by the warehouse supervision system thanks
to heuristic models and algorithms for the correct scheduling of the missions
assigned to the AGV vehicles.
An AGV system must deal with variable and dynamic executive conditions.
The operating flow entering the system is typically irregular and unpre-
dictable, certain areas of the warehouse can be temporarily closed and the
AGVs themselves can enter and exit the system because of maintenance. Al-
though these work situations may arise, the system must operate in a robust,
efficient and safe mode.
In conclusion, the realisation of a material handling system using AGV vehi-
cles offers advantages such as: operational flexibility, reduction of the overall
handling time, better control of the process and the materials being worked
on and balancing of production with better quality.
During the planning stage it is necessary to guarantee a correct labour dis-
tribution among the various systems that allow the overall management of
the automatic warehouse.
The typical topology in automated warehouses with AGV vehicles consists
of:

1. Warehouse management system WMS: it has the task of supporting

the operational management of the physical flows passing through the
warehouse, from the control of incoming goods during the acceptance
phase to the preparation of shipments to customers;

2. Warehouse control system WCS, also called supervision system: it is a

software solution with integrated control and real-time monitoring, that
manages the flows of goods during their movements on different types of
automatic equipment. It is able to exchange ongoing communications,
process commands and signals and optimise the flow of materials via
strategic decisions;

3. Master PLC or central calculator: ensures the real-time management

of a set of devices and acts as an intermediary between the supervision
system, typically located in servers away from the warehouse, and the
on-site machines;

4. PLC inside every single device.

21
 2. RGV system:
2.2. RGV - Rail Guided Vehicle system design and control

A proper implementation and coordination between the presented systems

enables the pre-established objectives to be met, and the advantages that
AGV vehicle technology makes available to be fully exploited.
Efficient and effective solutions should be sought by adopting heuristic ap-
proaches in order to obtain the right compromise between calculation time
and quality.

2.2 RGV - Rail Guided Vehicle system

The routing systems are divided between those with a fixed layout and those
with a variable layout. The first require the installation of a track on the
floor - for example a rail or magnetic belt - or underneath the surface; the
second require a map of the area in which the vehicles operate and a series
of fixed reference points that can be tagged. Certainly, the second category
allows greater flexibility, but requires superior computation capacity on the
part of the central calculator; it can be said that with the routing systems of
the first type the vehicles follow a path, while with the systems of the second
type they choose a path.
In AGV systems, the most common vehicles are those with a mechanical
guide: the electrically-driven transport vehicles with on-board calculator
move along specific paths, guaranteeing high degree of precision both when
driving and approaching the stations.
The RGV system is a special transport system on an electrified monorail,
very common in all industrial sectors where logistic efficiency is a key factor;
in many production processes and in the connection between different areas
of the warehouse, it is in fact particularly suitable and convenient for routes
that cover long distances.
It is a relatively simple transport system, where, in the traditional version,
the circuit is composed of an electrified aluminium belt with graphite and
copper brushes that run over it. These brushes guarantee the power required
for the power supply and the independent movement of each machine.
The rail is hung from the ceiling to reduce clutter on the floor, and is con-
figured in relation to the space and the characteristics of the building, which
consists of straight lines, curves, turnouts and parallel lines. The result is a
versatile and flexible transport system that allows future modifications and
expansions. The circuit, if necessary, can be redesigned and new vehicles
can be inserted at any time. Typically, the number of vehicles is chosen in

22
 2. RGV system:
2.2. RGV - Rail Guided Vehicle system design and control

Figure 2.1: RGV vehicle

relation to the quantity of pallets per hour that they wish to move. A cor-
rect decision requires an in-depth analysis and the use of specific indicators
such as the lead time and the potential handling presented in the previous
chapter.
The RGV system ensures an increase in productivity and a reduction in the
space required. Transportation is guaranteed at high speed and performance
levels. The maximum speed of an empty vehicles is 2000 mm/sec, which is
reduced in relation to the load they are delivering or on particular sections
of the route. This system is considerably faster than the most common con-
veyor systems, where the speed is around 500 mm/sec. Furthermore, the
space and time required to move the pallets between the different stations is
minimal.
The vehicles, which include chains or rollers, can move products of different
sizes, formats and weights. Usually intended for the transport of pallets,
they are suitable for types such as euro-pallets (1200 x 800 mm) and amer-
ican pallets (1220mm x 1020mm). They are able to transport goods with a
maximum weight of two tons, whether individual or composite pallets.
The RGV system has advantages and disadvantages compared to other types
of AGV vehicles; the choice is a decision that is up to the system designer.
It guarantees a very high level of safety. The operations are completed with
completely automatic procedures and under the supervision of the central
system. The products are handled in a silent, clean and efficient manner.

23
2.3. Guide-path design: 2. RGV system:
single loop type design and control

2.3 Guide-path design:

single loop type
The efficiency of an RGV system is strongly influenced by the path designa-
tion algorithm and it means that it needs a careful planning; in this regard,
it is considered the first problem to be evaluated.
The path depends mainly on the space and layout of the plant, which estab-
lishes physical and fixed constraints on the problem. Large spaces with high
manoeuvring possibilities increase the number of admissible routes, reduce
travel times and guarantee greater decision-making capacity for the central
system in relation to the current handling flow.
The route has a significant effect on the time it takes for vehicles to perform
tasks. The more alternatives and intersections there are, the more compli-
cated it becomes to control the traffic in an efficient way.
Each layout can be represented by a graph G(N, A), where N is the set of
nodes that evidences the operating stations and A is the set of arcs that de-
fines the connecting sections between adjacent nodes. Each node has at least
two incident arcs, for incoming and outgoing movement respectively. The
operating stations are locations in which it is possible to carry out picking
and depositing operations and provide for a temporary release of the product
pending the progress of sales or storage operations.
The arc direction indicates the vehicles direction flow and in each arc can be
assigned a cost representing the distance between the designated nodes or
the time required to travel itself.
In the literature various models have been presented, among which the most
commonly used one has as objective function the minimisation of the dis-
tance travelled considering both transfers, when the vehicle is loaded and
when it is empty (Kaspi et al, 2002).
Among the EMS rail systems, there are three main configurations when deter-
mining the route, presented by the literature. The literature does not explain
where and when it is appropriate to use one specific design over another, the
decision must be taken in relation to the requirements and characteristics of
the system.
The conventional driving system is composed of a network of routes that
connects all the stations, through straight lines, curves, intersections and
shortcuts. The conventional system can be unidirectional or bidirectional
in relation to needs and the number of rails. They involve a high level of

24
2.3. Guide-path design: 2. RGV system:
single loop type design and control

complexity, both computationally and in terms of decision-making, with con-

siderable control and efficient difficulties.

Figure 2.2: Conventional guide path system

The tandem configuration is based on the strategy of splitting frequently used

routes into zones. It basically divides the entire route into non-overlapping
circuits, where each loop is served by a single vehicle. The main advantage is
the reduction of congestion, blocks and interference due to high traffic, but
it restricts the number of vehicles to one per zone and involves the creation
of innumerable restricted areas to satisfy large flows of movement.
The main disadvantage is the need to declare a zone inactive when its re-
spective vehicle is inoperable due to malfunction.

Figure 2.3: Tandem guide path system

In conclusion, the single-loop system is composed of a single path with-

out any branches, in one direction and served by one or more vehicles. Bi-
directionality in these cases is possible but involves possible interference be-
tween vehicles. This system ensures simplicity of implementation and op-
eration with respect to the previous types, while considerably reducing the
flexibility of the system in the face of problems such as vehicle malfunctions,

25
2.3. Guide-path design: 2. RGV system:
single loop type design and control

which involve the shutdown of the entire system while awaiting the action of
an operator who allows the vehicle to exit the circuit, typically transporting
it to a branch set up specifically for maintenance.

Figure 2.4: Single loop system

The production capacity of a single loop system is considerably reduced when

compared with conventional route systems; in this regard a higher number
of vehicles is required to obtain acceptable results. The characteristics of the
single loop make it possible to avoid the problems of intersections on the path
and considerably reduce the computational complexity, and consequently the
final cost.
Within a single loop a barcode string, a subject discussed in greater depth in
the following sections, determines the position of the vehicles. The barcode
value is unique and incremental throughout the path, but, being a circuit,
there is a junction point between the end and the start that involves a jump
in the reading value. In this regard, to allow correct policies for determining
position, for calculating the distance to be travelled and for process the ve-
hicle’s journey, it is convenient split the circuit into sub-sections.
A section is defined as a range of contiguous barcode values with start and end
values, such that the constraint start value ¡ end value is always respected.
Partitioning the route into sections implies the following advantages:

1. an increase in the safety distance between vehicles can be considered

in curvilinear sections, where the vehicle encumbrance is greater;

2. the avoidance of queuing in determined zones, such as handling points

for unloading and loading of the products, points of exchange with the
maintenance branch etc;

26
2.3. Guide-path design: 2. RGV system:
single loop type design and control

3. the maximum speed is increased or reduced in relation to the type

of section - rectilinear zones allow higher speeds than others - thus
adapting the speed of the vehicles to the physical characteristics of the
path.

The route structure is then composed by a set of sections, modelled as follows:

Figure 2.5: Section data type

The top three fields allow to generate a unique chain of sections, while the
rest concern the topics discussed above.
In single loop systems, the journey - starting from a specific node to reach
another one - is unique and it is not necessary to determine the path to
take. The cost of the arc is calculated as the difference in position between
end node and initial node, measurable using the Compute distance to target
position algorithm, which takes into consideration the sections model.
During communication with the vehicle, the central calculator dynamically
updates the distance that the vehicle must perform to reach the final position,
which can be a node or the maximum distance that can be travelled due to
the vehicles physical presence along the route. Since there are no crossings
or other viable routes, the vehicle continues its movement in a straight line.
Until the designated node is reached, where the missing distance will be zero.

27
2.3. Guide-path design: 2. RGV system:
single loop type design and control

2.3.1 Compute distance to target position algorithm

The following algorithm determines the distance between two different po-
sitions. The algorithm is performed by the vehicle’s internal calculator and
used in various functions such as speed calculation.

Algorithm 1: Compute distance to target position

Input : sections, current position [mm], target position [mm];
Output: distance [mm];
1 current section = Get section from position (current position);
2 target section = Get section from position (target position);
3 if current section is equal to target section &&
current position ≤ target position then
result = target position - current position;
else
result = end position value from current section - current
position;
4 foreach section do
next section = Get next section from ID(id from current section);
if next section is equal to target section then
result = result +
(target position - start position value from next section);
exit;
else
result = result +
(end position value from next section - start position value
from next section);

28
2.3. Guide-path design: 2. RGV system:
single loop type design and control

2.3.2 Get section from position algorithm

The following algorithm determines the section to which a given position
belongs. To do this it uses the property of uniqueness of the position value
within the route.

Algorithm 2: Get section from position

Input : sections, position [mm];
Output: section;
1 foreach section do
if start position value from section ≤ position ≤ end position
value from section then
result = section

29
2.4. Communication 2. RGV system:
between entities design and control

2.4 Communication
between entities
The need to be able to transmit data remotely, in real time, occurs in many
situations and in areas that are very different from each other; in fact, often,
the place where data is processed and managed is far away from the system
that must execute them.
Communication in automated plants plays a fundamental role given the high
quantity of information. An example of this is represented by the data relat-
ing to the mapping of the warehouse or the handling area, the information
on the products and the fleet management in EMS systems.
In EMS systems, communication is bidirectional: according to the configura-
tion of the vehicles, the vehicle contains one or more fixed antennas, which,
connected to the internal calculator, guarantee the reception of information;
and a transmitter, through which it sends the data to the system concerning
its identity and status.
The central calculator controls and coordinates the fleet of vehicles and the
remaining system components, such as working stations, conveyors and ele-
vators, via an ethernet-type wired network.
The architecture most used for WLAN communication in EMS systems is
called ”access point based”, in which, through one or more access points,
it is possible to connect wireless equipment to a LAN network, conveying
information on a wired support. It allows you to connect wireless mobile
equipment to an existing fixed network, extending its connectivity. The
main WLAN technologies available are infrared and radio frequency. Ra-
dio frequency transmission is much more used because, unlike with infrared,
sending and receiving is possible even if the transmitter and the receiver are
not visible, which is a typical circumstance in many operating conditions.
The type of frequency signal used is the broadband radio signal, so that,
thanks to the European legislation ETSI 300-220, transmitting equipment
can be used without the need for any license from the Ministry of Telecom-
munications. These so-called low power devices must transmit 0.01 W at
low power, which also implies a reduction, practically zero, of the risks for
the health of the operators; just think that a mobile phone transmits with
a power of 2 W at 900 MHz. However, the low power limits the range of
action, which goes from a minimum of a few tens of metres indoors to a max-
imum of around a hundred metres outdoors, in practice covering a thousand

30
2.4. Communication 2. RGV system:
between entities design and control

square meters. However, in many cases, if the area to operate is greater, it is

possible to position the various repeaters extending the coverage as desired
with the only limit set by the management capacity of the software used.

Figure 2.6: Communication topology

The main data transmitted, unavoidable for correct system and vehicle man-
agement, are:
1. the enabling: signal that enables the vehicle to move and whose use-
fulness comes into play when anomalous or emergency situations arise,
and an immediate stop is required;

2. the route mapping: represents the sequence of the sections of the route.
Saved and configured in the central calculator, it is sent to the vehicle
upon specific request or when the model has been modified;

3. the destination value: indicates the position that the vehicle must reach
- it can be a destination node or a momentary value due to the physical
presence of other vehicles on the path;

4. the current position: the diffusion of the signal allows the central system
to know in real time the position of each vehicle, avoiding collisions and
congestion;

5. the present mission: represents the knowledge, by the vehicle, of having

a mission assigned or not, since a movement can also be a request for
removal from the current path;

6. the state of the sensors: identification of the state of the motor, of the
barcode reader and of the photocells that detect physical presence. In
order to find mechanical anomalies and damage of the components;

31
2.4. Communication 2. RGV system:
between entities design and control

7. the loading and unloading permit: the central system, coordinating

with the vehicles, allows the synchronisation of operations with the
workstations, assigning and removing execution permits;

8. the keep alive bit: the parameter defined ”keep alive” is of fundamental
importance in determining network problems. It allows you to identify
a non-explicit disconnection by the vehicle, making sure that the con-
nection with the system is still active and working. A vehicle could be
considered stationary in a given position since no updates are sent on
its displacement, but on the contrary, physically the vehicle is moving.
To recognise these emergency situations, where the logical represen-
tation differs from the physical one, the keep alive parameter makes
it possible to determine, through a continuous exchange of signals, if
there is a fault on the network, which causes delays in transmitting
or receiving data. A delay in position reception is not allowed and
can lead to collisions and subsequently requiring massive maintenance
interventions.

The mission is a specific material transport task, including operations of

retrieval and deposit from defined sources and to prescribed destinations. It
does not represent a parameter to be sent, because it is composed of the
destination value and the loading and unloading permit. Only the central
calculator and the supervisor system need information about the status of a
mission, while the operating vehicle is not responsible for it.
In the case of single-loop EMS systems, the route to reach a destination node
is also a piece of information whose transmission is not required.
The same communication principles seen previously are reproduced for data
transmission between the central calculator and the warehouse supervision
system.
It is a duty of the supervisor to determine the priority of the missions to be
carried out, the designation of the missions to the vehicles and the timing
of entry and exit of goods in relation to the actual and future forecasts and
then to communicate them to the central calculator. The information can
come in real time from the plant, from others systems like the warehouse
management system or from its own knowledge of the current progress of
customer orders. It requires the information from various systems to make
decisions that are best suited to optimising the flow.
Communication is therefore fundamental for the fleet control and the entire

32
 2. RGV system:
2.5. Material handling design and control

system supervision, so as to instantly identify the correct arrangement in the

succession of events and communicate it to the appropriate entity.

2.5 Material handling

Material handling is the process of transferring products from a source loca-
tion to a destination. RGV vehicles have the task of positioning themselves
in front of the working station, coordinating with the central system so that
the picking operation takes place correctly and then moving towards the des-
tination, carrying out the dual process.
The systems involved are the central calculator, which manages and con-
trols the stations and the plant points, and the individual vehicles, which
are responsible for moving the goods. Coordination and synchronisation are
required during picking and delivering, both to ensure the right conclusion
of the operation and to prevent alerts or anomalies.
The calculator must in fact recognise failed conclusion whether on the station
side and on the vehicle side. The procedure takes place via two handshake
signals, the handling permission and the handling in progress.
The respective entities recognise an anomalous situation in relation to the
received handshake signals, which differ from the expected behaviour.
If either of the two entities behaves differently than expected, an alert is
signalled during the transfer cycle and a maintenance phase is requested for
the operators.
An approximate time is also indicated for the operation conclusion to avoid
deadlock situations; if the times are not respected, an error message is sent
to the supervisor system. So that the steps of each operation are executed
and tracked, a finite state machine is used, which makes it possible to keep
track of both, the steps and the errors, in which they occurred.
Each operation is preceded by three preconditions that must be respected in
order to start the requested operation, material loading or unloading.

2.5.1 Loading and unloading preconditions

The preconditions differ in relation to the entities involved; they are a form
of communication between two separately managed systems, so that control
criteria are required both on the central calculator side and on the vehicle
side. The conditions shown represent the goods loading operations; the un-

33
 2. RGV system:
2.5. Material handling design and control

loading phase has similar constraints, but with a dual procedure.

The physical conditions of the central calculator determine the physical pres-
ence of products in a load bay and it is waiting to be picked up by a vehicle.

1 physical condition =
physical presence in working station &&
obstruction not intercepted &&
vehicle is empty &&
vehicle has reached the target position;

The logical conditions of the central calculator determine the logical presence
of the products in the loading bays. This value keep track of the progress
made and guarantee the constant control of the movement.

1 logical condition =
mission type is equal to load operation &&
vehicle is not in error mode &&
permitted signal from vehicle = 0 &&
rolling in progress signal from vehicle = 0 &&
vehicle has not logical presence &&
working station has logical presence;

The central calculator cycle conditions verify that there is no cycle in progress
in adjacent system points and that no errors have occurred previously.

1 cycle condition =
current cycle is not executing &&
current cycle has not fault &&
next cycle is not executing &&
next cycle has not fault;

34
 2. RGV system:
2.5. Material handling design and control

The vehicle’s preconditions require fewer controls and are easier to calculate.

1 physical condition =
vehicle position is equal to working station position &&
vehicle is empty;

2 cycle condition =
current cycle is not executing &&
current cycle has not fault;

2.5.2 Handshake
As described above, in order to guarantee a correct communication between
the systems involved it is required the usage of two signals: handling permis-
sion and handling in progress.
The first value represents a question to which corresponds a positive or neg-
ative answer, the second indicates the actual movement of the rollers by the
corresponding entity.
A single signal does not allow for correct interpretation at the action end. If
only the operation in progress signal is used, its conclusion could be due to
either successful completion of the transfer or to shutdown due to anomalies.
On the other hand, with the introduction of at least two signals it is possible
to figure out two or more different irregularities.
The handshake phase always starts from the central calculator, which has:
direct contact with the WCS, the complete plant management and a greater
control of general operating; while the subject who completes the ending
phase changes in relation to the action carried out.
Their behaviour is shown by the following images, first for the loading and
then for the unloading procedure.
In the loading procedure the goods are moved from the plant point to the ve-
hicle. The signals permitted! and rolling in progress are sent simultaneously
from the vehicle to the calculator; ideally the one who receives the product
moves its rollers before the source. This sequence means the goods will not
get turned around and bent during the transfer.
To advise that the transfer has taken place correctly, the vehicle then resets

35
 2. RGV system:
2.5. Material handling design and control

Figure 2.7: Handshake in load cycle

both signals and waits for the reply in order to consider the operation suc-
cessfully completed. Upon receiving permitted! = 0 and rolling in progress =
0 in the same clock cycle, the calculator checks its own status and responds
with the conclusion of the operation. At this point the vehicle is loaded and
ready to move to a destination node.

Figure 2.8: Handshake in unload cycle

The unloading procedure involves the transfer of the goods from the vehicle
to the designated plant point. The handshake is very similar to the previous
one; where the central system is the initiator of the operation and the vehicle
the receiver.
In the current circumstances, the vehicle awaits the movement of the rollers
on the part of the station, which appears as the product destination and,

36
 2. RGV system:
2.5. Material handling design and control

only after receiving the signal unload in progress = 1, the transfer begins,
signalled by rolling in progress = 1.
The central system advises the vehicle that the handling has come about
correctly by sending signals unload permitted?? = 0 and unload in progress
= 0. The conclusion is subsequently brought about with a positive response
from the vehicle.

2.5.3 Finite state machine

Loading and unloading operations do not only involve the affected systems,
but also a set of sensors, inverter and motors, such as photocells and position
control devices. To ensure supervision and management of the transfer, sim-
ple finite-state machines are built for the central calculator and the vehicle,
where the chain of steps represents the correct sequence of actions up to the
conclusion.
The states are represented by circles and named according to the step, while
the transactions represent the occurrence of definite events. The starting
states are named idle, while the conclusion states are M6 and T6.
The finite state machines for loading procedures is shown below and the
states of the central calculator are explained in-depth.

37
 2. RGV system:
2.5. Material handling design and control

Figure 2.9: Finite state machines for loading procedures

38
 2. RGV system:
2.5. Material handling design and control

In state M1 the two systems begin the handshake phase, the product is
physically in the loading area, indicated by the photocell that detects phys-
ical presence, and it is waiting to be transferred. There is also a second
photocell, called obstruction or anti-fall, which checks that there are no ob-
structions before, during and after the movement between the station and
the vehicle.

1 M: load permission?? = 1 & load = 0;

2 T: permitted! = 0 & rolling in progress = 0;
3 FTC presence in working station = 1;
4 FTC obstruction = 0;
5 Timer = 0;

In state M2 the handshake phase continues, the vehicle has consented to

the load request by the central calculator and has started moving the rollers
whilst waiting to receive the product. The goods are still physically at the
plant point, awaiting the activation of the rollers.

1 M: load permission?? = 1 & load = 0;

2 T: permitted! = 1 & rolling in progress = 1;
3 FTC presence in working station = 1;
4 FTC obstruction = 0;
5 Timer = 0;

In state M3 the first handshake phase ends, the central calculator moves the
rollers and the product transfer begins.

1 M: load permission?? = 1 & load = 1;

2 T: permitted! = 1 & rolling in progress = 1;
3 FTC presence in working station = 1;
4 FTC obstruction = 0;
5 Timer = start;

In state M4 the physical movement is completed, the product is on the ve-

hicle and the presence photocell of the station indicates the absence of the

39
 2. RGV system:
2.5. Material handling design and control

1 M: load permission?? = 1 & load = 1;

2 T: permitted! = 1 & rolling in progress = 1;
3 FTC presence in working station = 0;
4 FTC obstruction = 0;
5 Timer = stop;

goods.
In state M5 the second phase of the handshake began, with the task of advis-
ing the central calculator that the transfer has been completed successfully.
The vehicle resets its signals in the same clock cycle and waits for a response.
At this stage the product is physically inside the vehicle.

1 M: load permission?? = 1 & load = 1;

2 T: permitted! = 0 & rolling in progress = 0;
3 FTC presence in working station = 0;
4 FTC obstruction = 0;
5 Timer = 0;

The last state M6 concludes the procedure and the second phase of the hand-
shake. The central calculator has checked that the obstruction photocell does
not present anomalies and afterwards resets its own signals. At this reception
the vehicle considers the loading phase correctly completed and starts transit
to the designated destination node.

1 M: load permission?? = 0 & load = 0;

2 T: permitted! = 0 & rolling in progress = 0;
3 FTC presence in working station = 0;
4 FTC obstruction = 0;
5 Timer = 0;

State M10 occurs when the goods handling timeout is engaged; the time
taken for the transfer is higher than standard one and a situation of discrep-
ancy is announced. The central calculator reports the status to the vehicle
and stops the procedure.

40
 2. RGV system:
2.5. Material handling design and control

1 M: load permission?? = 1 & load = 0;

2 T: permitted! = 1 & rolling in progress = 1;
3 FTC presence in working station = 0;
4 FTC obstruction = 0;
5 Timer = 0;

State M11 implies an unexpected reception of the handshake signals. The

ensuing state involves an anomaly and requires the assistance of operators
for verification.

1 M: load permission?? = 1 & load = 0;

2 T: permitted! = 0 & rolling in progress = 1;
or
T: permitted! = 1 & rolling in progress = 0;
3 FTC presence in working station = 0;
4 FTC obstruction = 0;
5 Timer = 0;

In state M12 the procedure has not completed correctly; the obstacle photo-
cell has intercepted an anomaly and blocked progress. Upon the occurrence
of this irregularity, a notification is sent to the supervisor system and the
maintenance intervention by the operators is awaited. The calculator alerts
the vehicle by resetting only the load signal, which recognises the stall situ-
ation and stops the movement.

1 M: load permission?? = 1 & load = 0;

2 T: permitted! = 0 & rolling in progress = 0;
3 FTC presence in working station = 0;
4 FTC obstruction = 1;
5 Timer = 0;

41
 2. RGV system:
2.6. Idle vehicles design and control

2.6 Idle vehicles

Having vehicles without a mission is inevitable in an RGV system and, rather
than forcing them to return to parking areas, it is better planning strategic
points in the proximity of product release locations or loading bays.
There are two main strategies for the position of vehicles: static and dynamic.
In the static strategy the vehicle parking locations are selected to minimise
response time for a new movement request. Among them the distributed-
positioning rule uses multiple stopping points, which represent the stations
where it is possible to operate a loading action. The mission management
algorithm for idle vehicles is carried out by the central calculator which, in
relation to the state of the system and of each individual vehicle, assigns the
plant point closest to it.
When a vehicle becomes idle, the calculator assign a re-positioning mission
to it for the nearest stopping point, waiting for a subsequent task.
In the event that, during the journey, stationary vehicles are encountered,
the re-positioning algorithm makes a request for their movement owing to
the obstructed path. The same procedure is carried out when the selected
loading station is occupied by other idle vehicles. To which a new plant
point will be assigned. This strategy is valid if the number of stations is
considerably higher than the number of vehicles in the system.

42
 2. RGV system:
2.6. Idle vehicles design and control

2.6.1 Assign strategic missions algorithm

Algorithm 3: Assign strategic missions

Input : vehicles, missions, bays;
Output:
1 foreach vehicle do
if vehicle has a mission to do then
mission completed = false;
job mission =
vehicle has a mission to do && mission type is a job;
if job mission is false then
mission completed = true;
if mission completed is true then
station to stop = Get next closest bay where type is load;
mission type = re-positioning;
mission status = moving to destination;
mission destination = station position;
vehicle has a new mission assigned;

43
 2. RGV system:
2.7. Re-positioning management design and control

2.7 Re-positioning management

In EMS systems, displacement requests caused by the obstruction of the
route are a frequent factor due to the presence of a single track on which the
vehicles move. An idle vehicle placed in front of a loading bay, whilst await-
ing the assignment of a mission, could cause an obstacle for all subsequent
vehicles passing through that position. For this reason, a re-positioning re-
quest algorithm is essential for a correct system design.
The central calculator executes the algorithm, which is aware of the layout,
the vehicles position along the route and the missions agreed.
When the calculator recognises that the previous vehicle is blocked it as-
signs a blocking flag and checks the tasks assigned to the successive vehicles.
Re-positioning occurs when the vehicle has no mission assigned or when a
request for re-positioning is in progress but with insufficient distance in order
to clear the path.
The following algorithm reassign the position of all the idle vehicles which
hinder the journey of a mission, until arriving at the condition for which the
previous vehicle has a mission destination < mission destination from the
actual vehicle, so that the destination transmitted is further away than the
destination of the blocked vehicle.

44
 2. RGV system:
2.7. Re-positioning management design and control

2.7.1 Re-positioning request algorithm

Algorithm 4: Re-positioning request

Input : vehicles, bays;
Output:
1 foreach vehicle do
if previous vehicle is blocking && mission status is empty then
station to stop = Get next closest bay where type is load;
mission type = re-positioning;
mission status = moving to destination;
mission destination = station position;
vehicle has a new mission assigned;
if previous vehicle is blocking && mission type is re-positioning
then
if mission destination from previous vehicle ≥ mission
destination from vehicle then
new destination =
mission destination from previous vehicle + vehicle size;
station to stop = Get next closest bay where
type is load and position is equal to new destination;
mission destination = station position;

45
 2. RGV system:
2.8. Position design and control

2.8 Position
In relation to the type of AGV system, different technologies are used to
define the position: vehicles with laser triangulation use a laser device that
rotates 360 degrees and scans the surrounding area; vehicles equipped with
GPS adopt the differential technique with a fixed reference base in order to
accurately recognise the position; and finally, for the vehicles on tracks, po-
sition is established along the translation axis by a BCL sensor, positioned
on the driving wheel, which continuously reads a barcode string placed on
the track. The barcode is a system for transmitting information by storing it
in an optical figure composed of black and white lines, organised as parallel
bars and spaces with different thicknesses. The barcode can be defined as a
symbol for coding information in a format that can be acquired automati-
cally and cyclically by appropriate optical readers. The aim is to forward the
information in an automated processing system, guaranteeing its handling in
an easy, secure and economical manner.
In rail systems, the information contained in the bar code represents a nu-
merical value, such as the distance from the origin point. It involves an
ever-increasing value that is unique within the single-loop up to the exchange
point where there is a numerical jump to the origin point.

Figure 2.10: BLC sensor

In all barcode-based identification systems, four fundamental characteristics

must be taken into account: reliability, automation, precision and speed.

1. Reliability: for each reading operation, it is required to perform a vali-

dation action before considering the information correct. In RGV sys-
tems a wrong validated reading can lead to a vehicle logical position
different from the physical one. Which implies mistaken assessments
by the central calculator and consequent possible collisions.

46
 2. RGV system:
2.8. Position design and control

2. Automation: automatic and continuous data intake. Once installed

and configured the device performs the reading process continuously.

3. Accuracy: barcode-based systems guarantee a high level of precision.

In vehicles, the device reads from three to four codes simultaneously,
carries out the arithmetic average and returns the value. This proce-
dure reduces the error level due to worn or dirty strings.
The position of the device is also important to ensure accuracy in read-
ing. To read the barcode it is necessary to scan it along its entire length.
If it is too low, a slightly oblique scan, especially if done from a distance,
will not allow the optical reader to obtain the correct acquisition.

4. Speed: the reading operation takes place at each clock cycle of the
internal calculator of the vehicle, to the order of 250 scan/s.
As previously indicated, the decoding of the value detected by the barcode
reader is a fundamental part because it places the vehicle in a precise point
of the track. To detect any possible reading errors, it is possible to check - if
the path is one-way like in the single loop - that the new value sites in the
same section as the previous one or, at most, in the next one. When this
condition is not met, the vehicle reports a position that is not compatible
with its movement and it is signalled as an error situation where the vehicle
must stop and wait for confirmation or assistance by the operator.
Given that vehicles can only continue in one direction and that the route is
subdivided into sections as shown in the previous chapters, the reading value
during motion will increase within the same section.
This constraint guarantees better control over vehicle positioning. Indeed,
the following value must correspond to the previous one plus a delta value,
calculated in relation to the current speed of the vehicle. If the vehicle is
travelling on a section with high speed, the delta will be greater than a
stationary vehicle.
Assuming that at time t the vehicle is positioned at the validated position q
of the section t and that the validity interval is calculated by:

delta = actualspeed · clocktime [mm] (2.1)

actualbarcodevalue ≤ previousbarcodevalue + delta (2.2)

we have that:
actualbarcodevalue ∈ actualsection (2.3)

47
 2. RGV system:
2.8. Position design and control

or
actualbarcodevalue ∈ nextsection (2.4)

2.8.1 Barcode value validation algorithm

Algorithm 5: Barcode value validation

Input : sections, delta [mm], previous barcode value [mm], current
barcode value [mm];
Output: current position [mm];
1 current section = Get section from position (current barcode value);
2 upper bound = Add offset to position
(previous barcode value, delta);
3 lower bound = Subtract offset to position
(previous barcode value, delta);
4 upper bound section = Get section from position (upper bound);
5 lower bound section = Get section from position (lower bound);
6 foreach section do
if current section ID is equal to upper bound ID then
if start position of upper bound section ≤ current barcode
value ≤ upper bound then
result = true;
current position = current barcode value;
else
if lower bound ≤ current barcode value ≤ end position of
lower bound section then
result = true;
current position = current barcode value;

48
 2. RGV system:
2.9. Speed design and control

2.9 Speed
EMS vehicles, in order to travel the designated routes, need at least one elec-
tric motor, typically a first one for travel and a second one for moving rollers
and other components.
The inverter distributes power and voltage to the motor, assigns fixed pa-
rameters in relation to usage, and enables the movements. The vehicle’s
calculator interfaces directly with the inverter, communicating in order to
obtain information about the status of the motors and to control their speed,
expressed in revolutions per minute.
Consent to movement comes about via reception of the enabling signal from
the central system and the destination value that is different from the cur-
rent one, with the addition of a tolerance range. The vehicle moves according
to uniform rectilinear motion and uniformly accelerated rectilinear motion:
given the distance from the destination, the current speed and the maximum
speed of the current section, it is possible to calculate the set point for the
next time instant.
The UARM provides vehicle acceleration and deceleration with constant val-
ues. This motion allows the vehicle to follow a regular and predictable speed
profile. The maximum speed is calculated via three factors: type and char-
acteristics of the motor, transit speed in the current section and maximum
speed of the system, both expressed as a percentage.
Once reached, the vehicle will travel the section with uniform straight mo-
tion, excluding the presence of other vehicles or obstructions that require
deceleration.
The UARM equation:
1
x(t) = x0 + v 0 t + at2 (2.5)
2
v t = v 0 + at (2.6)
The unidirectional motion where speed
and acceleration are in the same direction:

v 2 − v 0 2 = 2a · (x − x0 ) (2.7)

Where:
· x(t) is the position at time t;

· x0 is the initial position;

49
 2. RGV system:
2.9. Speed design and control

· v0 is the initial speed;

· a is the acceleration constant.

The increment of speed can happen to a stationary vehicle or a vehicle enter-

ing a section where the maximum permitted speed is higher than the speed
allowed in the previous section. Acceleration takes place when the vehicle
has already entered the new section; on the contrary, in deceleration, the
vehicle must decrease its speed to have a maximum speed in the following
section that is not higher than the allowed one. This implies a deceleration
that takes place near to the new section, whilst the vehicle is still physically
present in the previous section. This deceleration situation occurs during
arrival at the destination node where the final speed must be zero.

Figure 2.11: UARM speed diagram

The operational data are: translation speed between 1500 and 2000 mm/s
in a straight line, between 1000 and 1500 mm/s in curvilinear section and
in junctions; acceleration and deceleration about 200 mm/s2, while the most
modern vehicles can have an acceleration of as much as 1000 mm/s2. Pallet
handling time, loading or unloading from 20 to 30 seconds.

50
 2. RGV system:
2.9. Speed design and control

2.9.1 Speed management algorithm

The speed management algorithm is calculated internally the vehicle, so as
to reduce the burden of computation to the central system and ensure greater
responsiveness. When the vehicle is near a section, on which a lower max-
imum speed than the current one is configured, the deceleration procedure
starts, so as to enter the section with an appropriate speed in regard to the
required characteristics.
The algorithm searches for the next sections within the vehicle stopping
space, calculated according to the UARM, evaluates the maximum speeds
configured and identifies the upper limit. If the current speed is lower than
the maximum speed allowed by the section and the vehicle is not decelerat-
ing, it is possible to accelerate by following the motion laws.

Algorithm 6: Speed management

Input : sections, system speed [%], motor parameters, acceleration
constant, distance to target [mm], previous set point [ mm
s
];
Output: next set point [ mm
s
];
1 current section maximum speed =
section speed system speed
maximum speed from motor parameters · 100
· 100
;
2 next section maximum speed =
maximum speed from motor parameters · next section
100
speed system speed
· 100
;
3 next speed square= target speed2
+ 2 · acceleration constant · distance to target;
√
4 next speed = 2 next speed square;
5 speed upper bound =
min(next speed, current section maximum speed);
6 speed lower bound = 0.0;
7 if previous set point < speed upper bound then
next set point = previous set point +
acceleration constant · clock100time ;
8 next set point = min(next set point, speed upper bound);
9 next set point = max(speed lower bound, next set point);
10 return next set point;

51
 2. RGV system:
2.9. Speed design and control

The variable target speed can assume two values: it sets 0.0 in case of vehicle
stopping at destination, or the previously calculated value by next section
maximum speed in the case in which the vehicle is closer a section with lower
permitted speed. The parameter distance to target represents the distance
between the current position and the destination node; calculated using the
algorithm 1: Compute distance to target position of guide path section. The
value next set point is secondly converted in revolutions per minute and sent
to the inverter that will take care of the direct communication with motors.

52
 2. RGV system:
2.9. Speed design and control

2.9.2 Speed management example

When a mission is assigned to the vehicle, whether it is a job or a re-
positioning request, the vehicle starts from standstill and then continues
with the acceleration phase until reaching the maximum allowed speed. The
presence of sections with different maximum speeds entails a stepped increase
in speed, as shown in the second graph. The second acceleration phase starts
only when the vehicle is completely within the section. The x-axis represents
the space, the y-axis represents the speed and the dotted line indicates the
beginning of a new section. According to the law of UARM, it takes 2.7

Figure 2.12: Acceleration phase example

1 start set point = 0.0 [ mm

s
];
mm
2 end set point = 1700 [ s ];
3 acceleration constant = 200 [ mm
s2
];
4 section 1:
speed = 60%;
distance required to reach the max speed = 2700 [mm];
5 section 2:
speed = 100%;
distance required to reach the max speed = 4600 [mm];

meters in section 1 to reach the maximum speed of 1020 mm/s through the
acceleration constant set; while in section 2 are needed 4.6 meters, starting

53
 2. RGV system:
2.9. Speed design and control

from 60% and reaching 100% of the speed. The speed percentage of section
1 puts an upper limit on the maximum speed so that the vehicle cannot
accelerate until the end of the section. When the vehicle has the task of po-
sitioning in a destination value, whether it is a loading or unloading node or
an obstruction due to the presence of other vehicles on the route, in relation
to its current speed it must start the deceleration phase with sufficient time
to stop, according to the law of motion. The speed reduction procedure for
the presence of a new section follows the same principles with the appro-
priate ratios. Deceleration follows a constant profile both for the stopping
procedure and for speed reduction near new sections.

Figure 2.13: Deceleration phase example

1 current set point = 1700 [ mm

s
];
mm
2 target set point = 0.0 [ s ];
3 deceleration constant = 200 [ mm
s2
];
4 section 1:
speed = 100%;
distance required to stop = 7300 [mm];

When a vehicle has to enter a section where the maximum allowed speed
is lower than the current speed, although being physically in the previous
section, the deceleration phase must begin in such a way as to arrive in the
next section at the required speed.
Once section 2 has been identified as the next one to be transited by the

54
 2. RGV system:
2.9. Speed design and control

1 current set point = 1700 [ mm

s
];
2 target set point = next section maximum speed;
3 deceleration constant = 200 [ mm
s2
];
4 section 1:
speed = 100%;
distance required to reach the new max speed = 5400 [mm];
5 section 2:
speed = 50%;

vehicle, and as having a maximum permitted speed lower than the current
one, the vehicle, using the speed management algorithm, at a distance of
5.4 meters from the beginning of the section starts the deceleration phase in
order to respect the established limit for the new section.

Figure 2.14: Multiple deceleration phase example

The same procedure is followed in the presence of contiguous sections with

different speeds, such as section 1 where the speed is 100%, section 2 at 50%
and, in conclusion, section 3 at 25%. The size of the sections is sufficiently
large such that the stop ramp is always guaranteed; indeed in section 2 the
vehicle continues in uniform straight motion up to 1,350 meters from the
beginning of section 3.

55
Chapter 3

Mission scheduler

Scheduling involves a methodologies family based on mathematical tech-

niques, which support the decision-making process and are used in many
problems, in particular in manufacturing industries for transport and logis-
tics. A scheduling problem consists, in its most general formulation, in the
optimal management of limited resources, designed to carry out certain activ-
ities while respecting appropriate constraints; the target is to evaluate usage
times of the resources and to assign material handling activities, called mis-
sions, to the individual AGVs; respecting constraints and following suitable
optimization criteria.
Once the hypotheses are defined and accepted, the scheduling problem can be
tackled and solved with a wide range of tools; in essence, these are algorithms
that provide possible solutions, bearing in mind the production context in
which they operate and the objective function to take into account.

56
3.1. The scheduler 3. Mission scheduler

3.1 The scheduler

The scheduling is a decision-making process frequently used in many indus-
tries in the manufacturing and services fields; it plays an important role in
improving production flow. It deals with the allocation of resources to tasks,
in defined time periods and with the aim of maximising/minimising one or
more parameters.
Resources and tasks can take different forms. Resources can be machines in
a work environment, processing units in a computer and so on. Similarly,
tasks can be operations in production processes, arrivals and departures in
airports, execution of programs in computing and others. Each task has def-
inite properties including the priority level, moment of release, time required
for execution and due date.
The objective functions can differ as well, from the minimisation of com-
pletion time of the last task to the maximisation of the number of tasks
completed before their respective deadlines.
Scheduling is a complex operation with the aim to establish the workload of
individual machines in the short term and to assign the planned missions.
Precisely because of this complexity, the objective that drives this operation
is not the costs minimisation, as the case for medium and long-term plan-
ning, but rather the optimization of the parameters that are representative
of the system performances.
Many companies use schedulers - software supports dedicated to the schedul-
ing phase - to process data relating to the distribution plans. Normally they
have a wide range of solutions and make it possible to vary in a multitude
of ways, both the objective functions and the methods - typically heuristic -
that are used for scheduling.
In a production context, the scheduler must interact with many other func-
tions located within the corporate IT system. Interactions are system-dependent
and can differ greatly from one situation to another. The local network of
terminals, workstations and machines, connected directly or indirectly to the
central system, is used to insert and to retrieve new information.
Ideally, the interaction should enable the quality of data and guarantee de-
tailed and complete tasks, which help the scheduler to find the solution and
maintain a high level of efficiency. Unfortunately, in real production environ-
ments, we rarely have tasks with rich information and, since the evolution
of the system is dynamic, the construction of a simple deterministic model
ends up being incorrect and inadequate.

57
3.2. Scheduling terminology 3. Mission scheduler

To this end, online models are used when there is no need to know, at time
zero, all the operations that must be scheduled, and they make it possible
to work with tasks with scant information. This is known as non-clairvoyant
scheduling and is more able to adapt to modern working environments.
It may not be immediately clear the impact that the scheduling has on the
objectives involved, but studies and practice have shown that the use of
a suitable scheduler is considerably important for the performance of the
system and it is opportune to dedicate time and effort to find an appro-
priate one. The realisation of a scheduling algorithm can be complicated
both from a technical and implementation point of view. The types of dif-
ficulties encountered on the technical side are the same as in other forms
of combinatorial optimization and stochastic models, while the difficulties
of implementation are a completely different kind - these could depend on
the accuracy of the model used for the analysis and the reliability of the
input data. Proper implementation requires time and research in order to
achieve reasonable efficiency and adequate control, reaching or approaching
the predetermined objectives.

3.2 Scheduling terminology

In all scheduling problems the number of tasks and resources is considered
finite. The number of tasks or operations is denoted by the letter n and the
number of resources or machines by m. Typically, the lower index j refers to
an operation while the index i refers to a machine.
If the execution requires a number of steps then the pair (i,j) refers to oper-
ation j performed on machine i.
The main tasks properties are:

· Execution time: pij

The field pij represents the time required to carry out operation j on
machine i. The index i can be omitted if the time does not depend on
the assigned machine or if only one machine can perform that specific
operation.

· Moment of release: rj
The moment of release of operation j, often called ready date or acti-
vation date, represents the instant in which the operation enters the
system and is ready to be computed.

58
3.2. Scheduling terminology 3. Mission scheduler

· Due date: dj
The due date of operation j represents its completion date or the com-
mitment that the company has made to the customer as regards the
conclusion of the order. The conclusion of a job, after its due date, is
allowed but incurs in penalty. Whereas when the due date is considered
mandatory and absolutely must be respected, it is called deadline.

· Weight : wj
The weight wj of the operation j is essentially a priority factor, which
denotes the importance of j with respect to the remaining operations
in the system. For example, the weight could represent the products
sequence during the loading phase in a truck.

As far as resources are concerned, different typologies are identified, such as

processing on: single machine, identical machines in parallel, machines in
parallel but with different speeds and so on. Among them, we examine in
detail only the presence of identical machines that carry out operations in
parallel, which will be the focus of this thesis and scheduler.

· Same machines in parallel: P m

In the system, there are m identical machines that are able to carry
out defined operations in parallel. Operation j requires a single process
and is assigned to a single machine i.

Some of the main features about machines are:

· Precedence constraints: prec

The precedents occur in environments with single or parallel machines
and require the conclusion of one or more operations before the others
may begin their execution. There are different forms of precedence:
if each operation has at least one predecessor and one successor, it is
called chain. If it only has a successor, we talk about in-tree whereas,
in conclusion, if it only has a predecessor, it is defined out-tree.

· Blocking: block
Blocking is an event that occurs mainly in processing flows. If the
resource has a limited buffer then, when they are all full, it may not
be possible assign new operations; they are placed on hold until a
subsequent space release.

59
3.2. Scheduling terminology 3. Mission scheduler

The objective functions differ considerably depending on the scheduler con-

text; many functions are in time domain and concern the completion time of
the tasks or of further operations to be performed subsequently. The com-
pletion time of operation j on machine i is denoted C ij . The instant in which
operation j leaves the system is defined as Cj .
Some examples of objective functions are:

· Makespan: C max
The makespan, defined as max(C1 , C2 , ... Cn ) is equivalento to the
completion time of the last operation to leave the system. A minimum
makespan value usually implies correct utilization of machines.

· Maximum Lateness: Lmax

The maximum lateness is defined as max(L1 , L2 , ... Ln ) and measures
the worst violation of the due dates.

Terminology makes a distinction between sequence and schedule. A sequence

usually corresponds to a simple permutation of the n operations or the order
in which operations are to be processed on a given machine.
The schedule refers to an allocation of operations within a more complicated
setting of machines, allowing multiple policies such as precedence, blocking
and more.
In recent years, a great deal of research on scheduling has been devoted to
discovery efficient algorithms. The results have shown that it is often not
possible to obtain optimal solutions in polynomial time; these problems are
called NP-hard. Confirming that a problem is NP-hard requires a formal
mathematical demonstration. Research has also focused on the boundary
line that distinguishes problems that can be solved in polynomial time and
in NP-hard solutions.

60
3.3. Online scheduling 3. Mission scheduler

3.3 Online scheduling

In the theory about main scheduling problems, it is assumed that the input
data are all known at time zero so as to allow the decision-making process to
determine the entire schedule, thanks to all the information at hand. This
standard paradigm is often referred to as offline scheduling.
In manufacturing environments, it is difficult to know or predict in advance
the flow of operations to be carried out, especially in contexts such as logistics
and transport where the handling material is irregular and unpredictable. In
this regard, the adoption of offline scheduling would not be possible, and a
new category of problems must be addressed.
In online scheduling the decision-maker does not know how many operations
are going to be processed and their respective properties. It becomes aware
of the existence of an operation only when it is released by the system and
presented to him, which can take place at any time during processing.
The scheduler must be frequently able to accept new operations and detect
the correct schedule with the information currently available. The online
scheduling problems seen in the literature are typically referred to online
over time scheduling problems.
A well-known algorithm for a simple resolution to online over time schedul-
ing problems goes by the name of LIST scheduling. The operations of the
decision-making process are initially ordered according to established criteria
and subsequently inserted into a list. Whenever the assignment of a task to a
resource must be considered, it checks the list and selects the next operation.
When a resource completes an operation, the decision-maker runs through
the list and assigns the next one. When an operation is issued and the pro-
cess has assigned the established priority to it, the operation may have to
wait until one or more resources become idle before being assigned and start
its processing. If the objective function is a regular performance measure,
then it may not take sense for the decision-maker to leave a machine idle
when there are still one or more operations on the list.
This procedure minimizes idle states in the machines and is perfectly able to
handle parallel machine problems, identifying a valid deal between simplicity
and precision of the solution.
Online scheduling in a parallel machine environment has received a signifi-
cant amount of attention during the last couple of years. Because it estab-
lishes a bridge between deterministic and stochastic scheduling. In stochastic
scheduling, decisions have to be made with high conditions of uncertainty

61
3.4. Types of algorithm -
the dispatching rules 3. Mission scheduler

and with a limited amount of information available. However the stochastic

paradigm is still quite different from the online paradigm; results show that
the solutions obtained in online scheduling often correspond to bounds in
stochastic scheduling.

3.4 Types of algorithm -

the dispatching rules
There are three main categories of scheduling algorithms:

· Exact algorithms can find optimal solutions;

· Approximation algorithms produce solutions that are very close to op-

timal and that are guaranteed to be within a fixed percentage of the
actual optimum. They are computationally fast and have polynomial
complexity;

· Heuristic algorithms obtain results which are not guaranteed to be close

to the optimum. Heuristic methods tend to be rather generic and can
be adapted to a wide variety of problems. The performance of there
methods is often evaluated empirically.

Figure 3.1: Types of algorithm

62
3.4. Types of algorithm -
the dispatching rules 3. Mission scheduler

Heuristic algorithms can be classified as construction heuristics or improve-

ment heuristics: in construction heuristics, the initial phase takes place with-
out a schedule and each time the system adds the operation entering the
system, whereas improvement heuristics begin with a schedule of the present
operations and, in a given time frame, try to find a better solution than the
current one, from the study of similar schedules.
Dispatching rules are an example of construction heuristics. Their main ad-
vantages are simplicity of implementation, even for particular cases, speed
of calculation and the possibility to find good solutions in relatively short
time frames. Whereas their disadvantages concern the identification of un-
predictable schedules when the rules that determine it are highly articulated.
Dispatching rules can be divided into basic rules and composite rules; com-
posites rules are obtained by the combination of two or more basic rules and
can guarantee significant improvements. The composite dispatching rules are
expressions that combine a number of n rules; each rule is ordered via a set
of specific measures. The result can be weighted by multiplication factors
with respect to the application context.
The ordering of the rules can be fixed, chosen by the designer through
decision-making constraints, or variable, in relation to special operations or
time functions.
Much research has been carried out in the last decade concerning dispatching
rules and notable rules have been studied in the literature. From a practical
point of view, the objective functions are considerably complicated, for ex-
ample a realistic one is given by the combination of different rules that are
fixed, variables and in the time domain.
In the following sections, three algorithms are presented which adopt dis-
patching rules as a scheduling strategy and which, via fixed rules, determine
the priority of transport missions in a logistical environment, before assigning
them to AGV vehicles for their execution.

63
3.5. Mission scheduler 3. Mission scheduler

3.5 Mission scheduler

The mission scheduler describes a set of algorithms aimed at determining the
correct sequence of operations to be performed, respecting complex logistical
criteria, and defining the assignment of missions to AGV vehicles.
The following scheduler, which adheres to dispatching rules theory, adopts a
set of fixed and ordered rules with regard to technical specifications received
from the transport management system TMS, in order to define the priority
of material handling missions during the input and output phase from auto-
mated warehouses equipped with autonomous guided vehicles AGV.
It is known that during the transport of goods, the number of clients that the
vehicle can satisfy typically ranges from 3 to 8 and it is necessary to respect
internal constraints regarding positioning of the products so that the driver
can unload the load in the shortest possible time.
The most commonly practice is the Last in First out policy, in which the last
product loaded onto the vehicle is the first to reach the designated destina-
tion, followed by the others. The data - which will become rules within the
scheduler - deriving from the transport management system include:
· Loading station: is the place where the truck is located - this also
corresponds to the destination point of the missions;

· Activation date: the date to which the operation of entry onto the
warehouse or sale of products corresponds;

· Vehicle’s daily number : a unique number to define the daily sequence

of the trucks;

· Internal position: the product position with respect to the truck.

The remaining useful data collected from other systems such as the company
management system, include:
· Movement type: the type could be an entry into the warehouse, an
internal movement or a sale of goods;

· Product: the product with the associated mission.

Combining these data enables a correct calculation of the priority value to
be assigned to each mission and subsequently to each vehicle. Missions are
created when a product requires a displacement, coinciding with a handling

64
3.5. Mission scheduler 3. Mission scheduler

phase within the warehouse. Reasons may vary, such as for example the
entry of new finished or semi-finished products for which the transportation
is required from acceptance station to internal location, or the sale of goods
for which displacement takes place from the warehouse location to the des-
tination bay for loading the truck.
In an environment where the regular flow of handling materials coming into
and leaving the warehouse is higher than 3000 units per day, it is essen-
tial to guarantee correct scheduling in order to satisfy transport constraints
and reduce waiting times. The priority value, once attributed to the cor-
responding mission, is used to establish the correct sequence of the goods
to be handled, and subsequently used by AGV vehicles to define which of
the products ready for loading has constraints that are more stringent than
others. The assignment algorithm checks the vehicles position on the circuit
and decides in relation to their status, which can satisfy the current mission.

Figure 3.2: Mission scheduler

65
3.5. Mission scheduler 3. Mission scheduler

Returning to the terminology treated in paragraph 3.2, the nomenclature

adopted for this scheduler corresponds to missions for the identification of
tasks and to vehicles for resources. The initial input for the sorting prod-
ucts algorithm is missionj and the final output from assignment algorithm
is given by the pair (missionj , vehiclei ).
The algorithms shown are implemented and processed in the warehouse con-
trol system WCS, the output are forwarded to the central calculator on site
and, in conclusion, sent as commands to the individual vehicles.

3.5.1 Mission data type

The mission entity, presented in the image, identifies a specific request for
movement within the warehouse. When the movement is completed, whether
correctly or not, the mission takes a completed status and is kept into the
database for prospective company performance studies. The mission and the
product item, guarantee the correct sequence of operations to be performed.
The fields that require greater analysis are:

Figure 3.3: Mission data type

· Mission type: makes it possible to distinguish between movements due

to sales or entry of products and movements due to obstructions along
the route. The values it can assume are job or re-positioning;
· Mission status: necessary to identify the current status of the mission
in the system; this field is of fundamental importance for following the
expected handling flow. The sequence shown, except for step 10, is the
one intended when the magazine is in automatic mode and no error

66
3.5. Mission scheduler 3. Mission scheduler

occurs during each operation.

The values that it can take are:

1. Undefined : when the mission is generated but its respective pri-

ority has not yet been calculated via the sorting algorithm;
2. Computed : the priority value has been calculated for the current
mission, but all the missions with the same destination do not yet
have an associated value;
3. Ready: the mission is waiting to be the next mission to execute;
4. To be assigned in vehicle: the current mission has priority over the
others and the mission scheduler calls on the assignment algorithm
to identify the best available vehicle;
5. Moving to source: the mission is assigned in vehicle and the phys-
ical movement is in progress, the empty vehicle is moving towards
the source of the product, which can be a plant entry point or a
location within the warehouse;
6. Load product: the loading cycle of the product onto the vehicle is
active. This procedure is described in section 2.5.
7. Moving to destination: the loaded vehicle is moving towards the
mission destination, whether a bay for loading onto truck or an
internal location;
8. Unload product: the last movement that the vehicles carry out
before completing the task, corresponds to the unloading of the
product at the destination. Also this procedure is described in
section 2.5.
9. Done: all the operations have been successfully completed and
the mission scheduler can undertake the next mission;
10. Error : shutdown due to anomaly.

All the missions contain the values for the source and destination fields, as
they are inserted during the generation of the mission, whereas vehicle ID is
added once the assignment algorithm has identified a vehicle capable of carry-
ing out the mission. The mission entity is essential for simple and controlled
management of the operations achieved by the automated warehouse.

67
3.5. Mission scheduler 3. Mission scheduler

3.5.2 Sorting products algorithm

The algorithm takes into account one mission at a time and, in relation to
those with the same destination station and with a previously defined prior-
ity value, it performs the calculation according to the TMS rules expressed
in the previous section.
Assuming that all the important data from the other systems has been cor-
rectly received, it is possible to determine the priority of the mission at hand.
The default priority value is indicated with X, meanwhile Y indicates the in-
crease or decrease index as regards the priority of the mission to be examined.
The algorithm is initially presented via a tree diagram and subsequently via
pseudo-code. The leaf node expresses the value to be applied to the priority
attribute.

Figure 3.4: Sorting algorithm

The assigned mission priority field corresponds to the weight property wj

68
3.5. Mission scheduler 3. Mission scheduler

analysed in the section concerning the terminology adopted in scheduler.

The X and Y parameters can assume any value as long as it is different from
zero when the sorting of sold products is required; indeed, if Y = 0 all the
missions take on the same priority.
On the other hand, in the shipment phase, Y must be sufficiently large as to
accommodate the generation of missions with activation dates much greater
than the current date if this is envisaged, so as to avoid priority values with
high number after the comma.
The pseudo-code concerning the sorting algorithm, the application of the tree
diagram of the previous figure, is shown below. It is assumed that the neces-
sary data are available and that missionj is the input value of the algorithm
so as to determine a correct output.
As appears from the code, the detection for what immediately precedes and
immediately follows the mission is extended when the result gives a negative
outcome, thus searching for an ever-wider set of solutions.
In the next chapter, we analyse the time frame required for the completion of
all the missions, and the computational cost of each algorithm. The sorting
algorithm takes relatively short time frames, considerably lower than those
required during the acquisition of data, obtained via database queries and
web service invocations to other systems.

69
3.5. Mission scheduler 3. Mission scheduler

Algorithm 7: Sorting products

Input : missions, missionj , destination bay, data from TMS,
default value X, interval Y;
Output: missionj priority;
1 missions with priority = Get all the missions where
status is equal to computed and destination is destination bay;

2 if missions with priority has at least one mission then

sorted missions with priority = Sort missions with priority by
increasing ACT IV AT ION DAT E,
then by T RU CK N U M BER,
then by ORDER IN T RU CK;

same mission = Search from sorted missions with priority

the first mission with
ACT IV AT ION DAT E equals to missionj activation date,
T RU CK N U M BER equals to missionj truck number,
ORDER IN T RU CK equals to missionj order in truck;

if same mission exists then

missionj priority = same mission priority;
else
previous mission = Search from sorted missions with priority
the last mission with mission priority;
next mission = Search from sorted missions with priority
the first mission with mission priority;
if previous mission exists && next mission not exists then
missionj priority = previous mission priority + Y;
else if previous mission not exists && next mission exists
then
missionj priority = next mission priority - Y;
else if previous mission exists && next mission exists then
missionj priority =
(previous mission priority + next mission priority)/2;
else
missionj priority = X;

70
3.5. Mission scheduler 3. Mission scheduler

Algorithm 8: Search from list the last mission with priority value
Input : missions, missionj ;
Output: previous mission;
1 Search from missions the last mission with
ACT IV AT ION DAT E equals to missionj activation date,
T RU CK N U M BER equals to missionj truck number,
ORDER IN T RU CK < missionj order in truck;
2 If it doesn’t exist then
Search from missions the last mission with
ACT IV AT ION DAT E equals to missionj activation date,
T RU CK N U M BER < missionj truck number;
3 If it doesn’t exist then
Search from missions the last mission with
ACT IV AT ION DAT E < missionj activation date;

Algorithm 9: Search from list the first mission with priority value
Input : missions, missionj ;
Output: next mission;
1 Search from missions the first mission with
ACT IV AT ION DAT E equals to missionj activation date,
T RU CK N U M BER equals to missionj truck number,
ORDER IN T RU CK > missionj order in truck;
2 If it doesn’t exist then
Search from missions the first mission with
ACT IV AT ION DAT E equals to missionj activation date,
T RU CK N U M BER > missionj truck number;
3 If it doesn’t exist then
Search from missions the first mission with
ACT IV AT ION DAT E > missionj activation date;

71
3.5. Mission scheduler 3. Mission scheduler

3.5.3 Mission worker

The Mission worker is the entity that keeps track of the mission status as
regards the destination point, distinguishing the missions as past, in progress
and future. Following the sorting algorithm, it takes from the database all the
missions ready for execution, subdivides them according to the destination
station, checks the last one executed and proceeds with the next one. This
algorithm has the task of determining to which mission, among those awaiting
assignment, it will first assign the resource once it becomes available. An
example is the case in which there are three missions ready for execution,
each with a different destination but respectively having maximum priority
within their set, and only two currently available vehicles.
The mission worker algorithm must choose which of the three missions assign
to the two vehicles, using criteria such as the type of operation - sales require
precedence with respect to entry flow to the warehouse - and truck waiting
time. This solution is based on the well-known Earliest Release Date first
policy in relation to the arrival of the transport vehicle in the warehouse.
This rule attempts to equalize the waiting times of the trucks to minimize
the variance of the waiting times. The respective objective function is:
X
V ar( (C j − rj )/n) (3.1)

The mission worker follows the List scheduling structure discussed in section
3.3, with some variations due to the application context and the presence of
parallel machines.
Missions to be executed is the matrix that contains all the missions to be
carried out and in progress, subdivided by destination and sorted by mis-
sion priority, columns and rows respectively. At each conclusion the worker
updates the matrix, always keeping the next missions to be scheduled in
first position. The current scheduler does not adopt a list to keep track of
operations but a matrix; in this regard, it is called a Matrix scheduling.

72
3.5. Mission scheduler 3. Mission scheduler

Algorithm 10: Search next mission to be executed

Input : missions, destination bays, data from TMS;
Output: next mission;
1 foreach destination bay do
mission in same destination = Get missions where status is
between ready and done and destination is equal to destination
bay;
sorted missions in same destination = Sort mission in same
destination by increasing mission priority;
missions to be executed = Insert sorted missions in same
destination;
2 next missions to be executed = Get first mission from each column
in missions to be executed where status is ready;
3 sorted next missions to be executed = Sort next missions to be
executed by increasing operation type and truck arrival time;
4 next mission to be executed = Get first mission from sorted next
missions to be executed;

73
3.5. Mission scheduler 3. Mission scheduler

When the missions are extracted from the matrix, those whose status is
in progress are not taken into account. The final result becomes the input
for the next algorithm, indeed next mission to be executed is the missionj
indicated as input for the assignment algorithm. The assignment of the mis-
sion happens via the assignment algorithm which has the task of identifying
the best AGV vehicle capable of satisfying the request.
The transition to the next mission takes place only after upon completion of
the previous one, the vehicle alerts the supervision system that the requested
operation has been completed and that it is waiting for a new movement to
carry out. The Mission worker simultaneously updates the mission pointer
and searches for the next one in the matrix.

3.5.4 Assignment algorithm

Once the calculation and identification phase of the mission is completed,
it is necessary to assign the movement to the vehicle in order to complete
the operation. The assignment algorithm determines, based on the euclidean
distance from the mission’s source point and other parameters shown below,
which is the most suitable vehicle for processing the task. Once it has been
identified, the mission is assigned, and the status of the Mission worker is
updated in mission moving to source point.
The current algorithm is affected by two main constraints of scheduling:
blocking and precedence, both analysed in the section 3.2. Indeed, the as-
signment is blocked if no vehicle is available for the execution of the selected
mission. Whereas precedence imposes the constraint that the mission must
be concluded before the assignment of the next one and the updating of the
matrix.
The search ends only when the vehicle is identified. In the event that there
are none, it is irrational to continue with other missions of different cate-
gories since there are no vehicles available. The completion of an operation
is awaited, before moving on to the assignment of the next mission.
The reset current mission in vehicle function removes the previous mission
assigned to the vehicle in order to give it the new one. The operations that
can be cancelled concern only the movements due to obstruction of the route;
the vehicle in the previous state was idle and was positioned in front of a
loading station according to the Assign strategic missions algorithm studied
in chapter 2. Subsequently a second vehicle, needing to pass through the
same section, due to the obstruction caused by the previous vehicle, has re-

74
3.5. Mission scheduler 3. Mission scheduler

quested its displacement via re-positioning. All movements of this type can
be cancelled because they do not involve real taks for the warehouse.

Algorithm 11: Assign mission to vehicle

Input : missionj , vehicles;
Output: pair(missionj ,vehiclei );
1 vehiclei = Get closest assignable vehicle with
position equals to missionj source position;
2 if vehiclei exists then
if vehiclei has already a mission && mission is re-positioning
request then
reset current mission in vehiclei ;
assign missionj to vehiclei ;

The algorithm output is given by the association missionj with vehiclei . In

the next chapter we will look at the calculation times required in different
vehicle availability conditions.

Get closest vehicle algorithm

The Get closest assignable vehicle function searches among all the vehicles
in the system for those who are currently not carrying out other movements
and who are able to stop in front of the mission source position. It excludes
vehicles that are undergoing maintenance, those that have other job type
missions in progress and those which have a load on board, whether physical
or logical.
Out of those remaining, it determines the interval between current position
and mission source point. The vehicle to which the mission is assigned is that
which has the minimum euclidean distance, and which is able to interrupt
its journey in the established location. An example is an idle vehicle moving
at maximum speed near to the mission origin. This may be the vehicle with
minimum distance, but it will not be able to stop with the correct decel-
eration ramp defined by the UARM and analysed in the previous chapter.
According to these considerations, the Get closest assignable vehicle function
does not consider this vehicle to be a valid solution and searches for the next

75
3.5. Mission scheduler 3. Mission scheduler

Algorithm 12: Get closest assignable vehicle

Input : missionj source position, vehicles;
Output: vehiclei ;
1 minimum distance = X;
2 foreach vehicle do
vehiclei = Get vehicle availability where
position is equal to missionj source position;
if vehiclei is not available then
continue;
actual distance = Get distance from
missionj source position and vehiclei ;
if actual distance ¡ minimum distance then
minimum distance = actual distance;
vehicle with minimum distance = vehiclei ;
3 return vehicle with minimum distance;

one.
The X value initially assigned to minimum distance represents a very high
value; a realistic example is the maximum distance between vehicle and sta-
tion. In the actual project it corresponds to the length of the circuit.
The Get vehicle availability function is the one which checks that the vehicle
is not carrying out other movements, that it is not under maintenance and
that the stopping time in the established location follows the deceleration
gradient envisaged by uniformly accelerated rectilinear motion.

76
3.5. Mission scheduler 3. Mission scheduler

3.5.5 Mission status updater

The Mission status updater function updates the mission status in relation
to the conclusion of the operation. The evolution of this field can be consid-
ered as a finite state automaton which makes it possible to describe mission
behaviour in a formal way. In what follows, only the steps concerning the
Mission scheduler are analysed, from the value undefined up to to be as-
signed in vehicle. The update of the last 4 values comes from the reception
of the operation completion data, received by the vehicle and by the central
calculator that manages the whole fleet of the system.

Algorithm 13: Mission status updater

Input : missions;
Output: missionj status;
1 foreach mission do
switch status do
case undefined do
invoke the sorting products algorithm;
if missionj priority has a value then
missionj status = computed;
case computed do
if all missions with same destination
have the priority value then
missionj status = ready;
case ready do
invoke the mission worker;
if missionj is the next mission to be executed then
missionj status = to be assigned in vehicle;
case to be assigned in vehicle do
invoke the assignment algorithm;
if missionj has a vehicle assigned then
missionj status = moving to source;

77
Chapter 4

Results

Results about the mission scheduler algorithms are analysed and commented
on below. In particular, we wish to determine computation times under
different handling conditions. In the analysis, we will ignore considerations
regarding the device hardware on which the program runs. We fully acknowl-
edge that computation times depend directly on the underlying hardware,
and that more powerful equipment will yield shorter times. The output, how-
ever, remain constant as a whole, respecting with appropriate proportions the
results and the graphs shown.

78
4.1. Sorting products algorithm 4. Results

4.1 Sorting products algorithm

The next table shows the calculation times, expressed in ms, for the sorting
algorithm. The content shall be interpreted as follows: given the mission j
as input, the algorithm uses N ms to calculate the priority value as output.

missions number computation time

mission 1 656
mission 2 974
mission 3 1364
mission 4 1010
mission 5 804
... ...
mission 150 781
mission 151 619
mission 152 972
... ...
mission 298 1574
mission 299 1071
mission 300 621

Table 4.1: Sorting algorithm results table

In the sections attached, the first graph shows all calculation times of the
algorithm. It is interesting to notice that, even though there are computation
differences between missions, the time remains constant until the last exe-
cution and that the interval is given by: 1863 – 555 = 1308 ms respectively
for maximum and minimum value detected. Deviations arise from calls to
the TMS service, which could introduce delays due to the telecommunica-
tions network. Still, delays are irrelevant relative to the total execution time,
which always tends to remain within a two-second interval.
Data collected date back to Monday 25/11/2019, where a total of 300 mis-
sions were prepared. The algorithm took about five minutes to determine all
priority values. It should be noted that the system is rarely required to cal-
culate a single, 300-missions block simultaneously; usually, requests comprise
60/70 missions, corresponding to the movement necessary to load a truck.

79
4.1. Sorting products algorithm 4. Results

Figure 4.1: Sorting algorithm computation time

As shown in the table, the time used remains constant for all mission
analysed. Therefore, the computational cost is polynomial, assumed O(n),
and follows the trend shown in the graph, which represents the first 50 cal-
culated missions.
The first mission should require the minimum calculation time since there are
no priority values in the database and the search for the immediately previous
and immediately next is unnecessary. Subsequent missions require a higher
number of calculations and searches but, even when the search expands its
range, the average time, required by the algorithm, remains unchanged; this
differences are not detected, and the algorithm shows similar times under all
conditions.

80
4.2. Mission worker 4. Results

4.2 Mission worker

The Mission worker, among the three models presented in the scheduler, is
the one which requires the minimum calculation time to determine the out-
put, the next mission to be perf ormed value. This is so because the infor-
mation it needs lies in memory, the algorithm is not subject to constraints,
differently than the assignment algorithm, and there is no need to make
queries or calls to external services. Among its main actions, we observe the
results for the construction of the scheduling matrix and for the ordering
and choice of the next mission to be performed. Results are broken down
into two cases: case 1 represents a real situation, in which there are 5 parallel
movements, sales or entries of products in the warehouse, and 10 missions for
each movement. Case 2 includes 10 movements and 50 missions, a situation
of high material flow. During the construction of the matrix, movements
represent the columns and missions the rows.
The next table shows the calculation times, expressed in ms, of the mission
worker. The content shall be interpreted as follows: given the case C, the
worker needs N ms to create the scheduling matrix and M ms to identify
the next mission to be perf ormed.

case matrix scheduling time next mission time

case 1 308 21
case 2 2275 34

Table 4.2: Mission worker results

Results show a calculation time that follows a polynomial trend, proportional

to the size of the matrix, indeed case 1 takes 0.3 seconds to build the matrix
and 0.02 seconds to choose the next mission, while case 2 takes 2.2 seconds
to build the matrix and 0.03 seconds to identify the mission.

81
4.3. Assignment algorithm 4. Results

4.3 Assignment algorithm

The following tables show the calculation times, expressed in ms, of the as-
signment algorithm. The content shall be interpreted as follows: given the
missionj as input, the algorithm uses N ms to produce the output (missionj ,
vehiclei ).

mission number vehicle ID assignment algorithm calculation time

mission 1 vehicle 5 1656
mission 2 vehicle 4 1344
mission 3 vehicle 3 1390
mission 4 vehicle 6 1781
mission 5 vehicle 7 1656
mission 6 vehicle 1 1953
mission 7 vehicle 8 1671
mission 8 vehicle 2 1453
mission 9 vehicle 5 1391
mission 10 vehicle 4 2130

Table 4.3: Assignment algorithm case 1

Case 1: dating back to Monday 25/11/2019, involves the realisation and

subsequent request for assignment of one mission at a time, with all vehicles
available for the execution. With this test, we can determine the real calcu-
lation times when there is at least one vehicle available for task execution.
Results show that the algorithm takes about 1.5 seconds per mission and the
output is smooth for all the input missions.
The case 1 looks at regular availability of vehicles in the system and it means
that the number of parallel operations is lower than the number of resources.

Case 2: dating back to Wednesday 27/11/2019, shows the activation of 10

parallel missions with 8 vehicles in the system, so the number of operations
exceeds the number of available resources. Accordingly, 10 − 8 = 2 tasks
will have to wait for at least two idle vehicles before their processing time.
Results indicate that the algorithm takes about 1.5 seconds from mission 1
until mission 8 to determine the output, scenario very similar to case 1. In
subsequent tasks, however, the situation changes considerably.
Task 9 features a high peak due to the maximum waiting time because of

82
4.3. Assignment algorithm 4. Results

mission number vehicle ID assignment algorithm calculation time

mission 1 vehicle 2 1296
mission 2 vehicle 6 1312
mission 3 vehicle 4 1359
mission 4 vehicle 5 1563
mission 5 vehicle 1 1921
mission 6 vehicle 3 1784
mission 7 vehicle 8 1478
mission 8 vehicle 7 1184
mission 9 vehicle 5 308895
mission 10 vehicle 4 91156
mission 11 vehicle 2 21468
mission 12 vehicle 6 17960
mission 13 vehicle 1 22653
mission 14 vehicle 8 30301
... ... ...
mission 30 vehicle 4 37031
mission 31 vehicle 3 42906
mission 32 vehicle 8 47520
mission 33 vehicle 1 33751

Table 4.4: Assignment algorithm case 2

the vehicles have recently and simultaneously started the execution of the
previous task. It follows that the algorithm fails to return an output until
a vehicle completes its tasks. This happens in about five minutes; this in-
dicates that, among the eight moving vehicles, that vehicle 4 is the first to
finish and took just over five minutes for its execution.
Task 10 shows a similar profile, with a reduced calculation time due to the
higher number of vehicles nearing the completion of their tasks. In fact,
the algorithm takes about one minute and half to calculate the output. By
adding this to the previous five minutes, we note that vehicle 2 completed the
task in six and a half minutes, which is very close to the average movement
time. These values arise from the blocking feature, analysed in paragraph
3.2, under which the algorithm operates.
In conclusion, we should note that, from task 30 on-wards, calculation times
concerning the wait for an available vehicle level off at 30-45 seconds for all

83
4.3. Assignment algorithm 4. Results

tasks. This phenomenon is due to a more uniform distribution of task dura-

tion, contrary to what happened in the first 15 seconds of this case.
The final table shows the results obtained by arithmetic average of cases 1
and 2, respectively for regular and poor availability of vehicles in the system.
The attachment 2 shows a graph about this scenario: the horizontal axis

vehicles availability computation time

regular 1642
low 40302

Table 4.5: Assignment algorithm average

indicates the time domain, the tasks assigned to vehicles appear under the
time axis and the arrows display the calculation time about each mission as
regard the vehicles availability.

84
4.4. Attachments 4. Results

4.4 Attachments
Attachment 1:

Figure 4.2: Sorting algorithm results

85
4.4. Attachments 4. Results

Attachment 2:

Figure 4.3: Assignment algorithm results

86
Conclusion

Loading or unloading a truck correctly is not just a question of safety: to

reduce transport prices, the number of required journeys and pollution, the
transport order preparation has recently become a real job for experts.
In this thesis project, a study was carried out on the logistics of the ware-
house, and a scheduler was implemented for the efficient management of tasks
assigned to RGV vehicles.
Three important criteria were taken into consideration when drawing up the
mission scheduler: the first concerns the quality of the solution obtained,
the second the amount of computation time to guarantee the solution and
the third the time needed to develop and implement heuristic algorithms.
Academic research typically evaluates only the first two aspects while, in an
industrial environment, the third is considered essential. In real plants, the
time required for implementing an algorithm must be restricted, and this is
one of the main reasons why, in practice, heuristics with local research are
more popular than more elaborate techniques.
The modules presented provide many advantages to the logistic processes.
The adoption of the sorting algorithm, which follows fixed rules for the load-
ing of products onto the vehicle, allows the driver - once he has reached the
respective customer - to reduce unloading times and the number of trips re-
quired. The product ordering sequence is a procedure carried out completely
automatically by the system, whereby it assigns the correct priority to the
missions. The use of these new techniques reduces the operator’s task just
to check up and supervision. The results showed that, from the computa-
tional point of view, the algorithm follows a polynomial trend. It is able to
determine the priority of 60 missions in approximately one minute.
The mission worker, the entity managing the ongoing missions, minimise the
waiting times gap of the different trucks, giving priority to sales movements
over entry movements, which typically occur from internal production facto-

87
4.4. Attachments 4. Results

ries. If a random solution is adopted, which does not optimise the process,
the consequence is that some trucks could wait a considerable amount of
time before the loading phase is completed. Its computational cost follows
a polynomial pattern and is relative to the number of parallel missions that
are performed. The calculation times are lower than the sorting algorithm,
the scheduling matrix requires about 300 ms and overall is able to handle
hundreds of missions in 50 ms.
Concluding with the assignment algorithm that identifies which, among the
available vehicles, ensures greater benefits for carrying out the current mis-
sion. Different choices would involve long waiting times for the products
before picking and consequently a delay in the average time of the entire
warehouse. The calculated time required when there is at least one vehicle
capable of satisfying the request is relatively low, between 1000 ms and 2000
ms. Otherwise, as analysed in the results chapter, there may be high peaks
in waiting times in relation to the current operations running on the vehicles,
with a maximum value equal to the longest movement – 1.
Personally, I am satisfied considering the scheduler and the results obtained.
We have succeeded in creating a product that satisfies all the features ex-
pressed by the customer, respecting the constraints and developing a model
able to identify valid solutions in reduced times. Given the regular movement
flow, the mission scheduler takes a maximum of 3 seconds from the initial
input point to the final output point.
Future developments include the introduction of new rules that will allow the
algorithm to identify better solutions than the current ones and to enhance
the dispatching rules adopted via new composite rules, for example given
by the combination of different basic rules and functions that consider the
waiting time in order to be executed.
An improvement would be researching a solution to avoid the precedence con-
straints in the assignment algorithm, which involves waiting for a completed
mission before the next execution. A valid alternative could authorise the
movement start but apply a check to ensure there are no previous missions
pending before the stage when the vehicle deposits at the final destination.
In conclusion, a further objective is to increase the set of vehicles valid for
a mission assignment. In this case, not only those who are not carrying out
any operations are considered but also those who will shortly complete a
mission close to the source of the next mission. This policy makes it possible
to extend the set of feasible solutions, reduce time and movements with idle
vehicles and improve warehouse efficiency.

88
4.4. Attachments 4. Results

The goal we are aiming for is an automatic warehouse that is effective, effi-
cient, simple and fast and that supports the internal activities of the company
and continuously improves customer service.
The development and testing of the RGV system, including the three algo-
rithms, discussed in chapters two and three, involved the Onit automation
team and me for about 8 months.
I would expressly thank the professor Silvano Martello, the professor Michele
Monaci and my colleagues. The first two for the support and the time they
gave me while I was working on this thesis, the latter for having involved me
in this challenging project and encouraged me at all times, working with you
is like being with a family. Thank you, everybody.

89
Bibliography

[O H97] C. Rajendran O. Holthaus. “Efficient dispatching rules for schedul-

ing in a job shop”. In: International journal of production eco-
nomics (1997).
[E F11] A. Regattieri E. Ferrari. Logistica integrata e flessibile. Esculapio,
2011. isbn: 9788874884643.
[R B12] S. Martello R. Burkard M. Dell’Amico. Assignment Problems -
Revised Reprint. Siam, 2012. isbn: 9781611972221.
[Mar14] Silvano Martello. Ricerca operativa. Esculapio, 2014. isbn: 9788874887965.
[Mar15] Angelo Renoldi Mara Bergamaschi. Logistica e supply chain man-
agement. Pearson, 2015. isbn: 9788865183663.
[16] “A new mathematical programming formulation of the Single-Picker-
Routing problem”. In: European journal of operational research
(2016).
[Pin16] Michael L. Pinedo. Scheduling - Theory, algorithms and systems.
Springer, 2016. isbn: 9783319265780.
[17] “Designing efficient order picking systems by combining planning
problems: State-of-art classification and review”. In: European jour-
nal of operational research (2017).
[Bal19] G. Balestri. Le basi della logistica. Hoepli, 2019. isbn: 9788820343354.
[D S] J. Tanchoco D. Sinriech. “The centroid projection method for lo-
cating pick-up and delivery stations in single loop AGV systems”.
In: Journal of manufacturing systems ().
[T L] M. de Koster T. Le-Anh. “A review of design and control of auto-
mated guided vehicle systems”. In: Rotterdam school of manage-
ment ().

90
List of Algorithms

1 Compute distance to target position . . . . . . . . . . . . . . . 28

2 Get section from position . . . . . . . . . . . . . . . . . . . . . 29
3 Assign strategic missions . . . . . . . . . . . . . . . . . . . . . 43
4 Re-positioning request . . . . . . . . . . . . . . . . . . . . . . . 45
5 Barcode value validation . . . . . . . . . . . . . . . . . . . . . . 48
6 Speed management . . . . . . . . . . . . . . . . . . . . . . . . . 51

7 Sorting products . . . . . . . . . . . . . . . . . . . . . . . . . . 70
8 Search from list the last mission with priority value . . . . . . . 71
9 Search from list the first mission with priority value . . . . . . . 71
10 Search next mission to be executed . . . . . . . . . . . . . . . . 73
11 Assign mission to vehicle . . . . . . . . . . . . . . . . . . . . . 75
12 Get closest assignable vehicle . . . . . . . . . . . . . . . . . . . 76
13 Mission status updater . . . . . . . . . . . . . . . . . . . . . . . 77

91
List of Figures

1.1 A supply chain example . . . . . . . . . . . . . . . . . . . . . 10

2.1 RGV vehicle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.2 Conventional guide path system . . . . . . . . . . . . . . . . . 25
2.3 Tandem guide path system . . . . . . . . . . . . . . . . . . . . 25
2.4 Single loop system . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5 Section data type . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Communication topology . . . . . . . . . . . . . . . . . . . . . 31
2.7 Handshake in load cycle . . . . . . . . . . . . . . . . . . . . . 36
2.8 Handshake in unload cycle . . . . . . . . . . . . . . . . . . . . 36
2.9 Finite state machines for loading procedures . . . . . . . . . . 38
2.10 BLC sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.11 UARM speed diagram . . . . . . . . . . . . . . . . . . . . . . 50
2.12 Acceleration phase example . . . . . . . . . . . . . . . . . . . 53
2.13 Deceleration phase example . . . . . . . . . . . . . . . . . . . 54
2.14 Multiple deceleration phase example . . . . . . . . . . . . . . 55

3.1 Types of algorithm . . . . . . . . . . . . . . . . . . . . . . . . 62

3.2 Mission scheduler . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3 Mission data type . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.4 Sorting algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 68

4.1 Sorting algorithm computation time . . . . . . . . . . . . . . . 80

4.2 Sorting algorithm results . . . . . . . . . . . . . . . . . . . . . 85
4.3 Assignment algorithm results . . . . . . . . . . . . . . . . . . 86

92
List of Tables

4.1 Sorting algorithm results table . . . . . . . . . . . . . . . . . . 79

4.2 Mission worker results . . . . . . . . . . . . . . . . . . . . . . 81
4.3 Assignment algorithm case 1 . . . . . . . . . . . . . . . . . . . 82
4.4 Assignment algorithm case 2 . . . . . . . . . . . . . . . . . . . 83
4.5 Assignment algorithm average . . . . . . . . . . . . . . . . . . 84

93