HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY

FACULTY OF MECHANICAL ENGINEERING

MECHATRONICS SYSTEM DESIGN PROJECT

Line Following Robot for Weight-Based Goods

Distribution
CLASS CC02 – HK251

Advisor: Trương Ngọc Cường

Group member: Bùi Hoàng Nhân - 2252557

Phan Minh Quân – 2252686
Nhâm Gia Nghĩa – 2212235
Đăng Lê Quang Nhật - 2252569
 CHƯƠNG 1: OVERVIEW

1.1 Overview of line-following delivery robot

1.1.1 Concept of line-following delivery robot

A line-following robot is an automated machine capable of following a predefined

path even when the path is changed by altering its shape. In the most basic case, the path
can be visible, such as a black line on a white surface (or vice versa), or invisible, such as
a magnetic field.
In addition to the line-following function, the robot is also designed with a loading
system for goods to be delivered. This system can detect whether goods have been
loaded, as well as distinguish the weight of the goods. From there, it delivers the goods to
the appropriate location based on the predefined weight.
A line-following delivery robot mainly consists of the following modules: power
supply, sensors, actuators, control system, communication between sensors and control
system, and communication between control system and actuators.

1.1.2 Operating principle of line-following delivery robot

Line-following is a task in which the robot must closely follow the line. The robot
must be able to detect a specific line and continue to track it. In addition, the robot must
detect the weight of the load placed on it and determine the path to the predetermined
location corresponding to the weight. When crossing intersections where multiple paths
are possible, the robot must determine the correct path according to the preset
requirements.
In this project, the robot will move along the drawn track on a given platform. The
robot follows the black line drawn on the white surface. Line sensors are used to detect
the line. When the infrared signal falls on the white surface, it is reflected; if it falls on the
black surface, it is only partially reflected. This principle is used to scan the track for the
robot. At the loading area, a package is placed manually on the robot. The robot detects it,
weighs it, and determines the path to the end location corresponding to the package’s
weight. Additionally, the robot must be able to turn at various degrees to follow curves
and must not be sensitive to environmental factors such as ambient light.
Typically, there are six types of basic movements for a line-following robot:
2
  Both left and right motors move forward → the robot moves forward. Conversely,
both in reverse → the robot moves backward.
 Left motor stops, right motor moves forward → the robot turns left. Conversely, it
turns right
 Left motor moves backward, right motor moves forward → the robot spins left.
Conversely, it spins right.

Table 1.1 Robot movement

Motor action
Robot
No.
movement
Left Right

1 Move forward Forward Forward

Move
2 Reverse Reverse
backward

3 Turn left Off Forward

4 Spin left Reverse Forward

5 Turn right Forward Off

6 Spin right Forward Reverse

3
 1.1.3 Products on the market applying line-following delivery robots

Autonomous Guided Vehicles (AGV):

AGVs are widely used in large industrial zones for automatic goods transportation.
They use photoelectric sensors for navigation and ultrasonic sensors to avoid collisions
that may affect goods and staff.
Advantages: long-term operation, ability to transport heavy loads. AGVs are
increasingly chosen to replace humans to reduce labor costs and increase accuracy.
AGV guidance technologies:
 Wire: Wires embedded in the floor transmit signals. AGV sensors detect these.
Limitation: inflexible, as routes are fixed.
 Guided tape: Magnetic tape or points stuck on the floor. They can be replaced
when routes change but may get dirty/damaged with heavy traffic.
 Laser target navigation: AGV uses laser transmitters/receivers with reflective
strips.
 Inertial navigation : AGVs follow floor-embedded signal markers controlled by
computer systems.
 Vision guidance: AGVs use 360° cameras to capture route features and build 3D
maps.
 Geo-guidance: AGVs detect columns, walls, and fixed objects to self-locate in
real-time and adjust paths without infrastructure limits.

Three main AGV types:

AGV Cart (Image 1.1) : The automatically guided cart is the most basic type of
AGV. Navigation systems can range from simple systems such as magnetic tape to
more complex path-guiding systems, based on sensors using AI for navigation.

4
 Image 1.1 AGV Cart
AGV Forklift (Image 1.2): The automatically guided forklift is an AGV designed to
function as a forklift. The difference is that it does not require a human operator and is
guided by a pre-installed navigation system.

Image 1.2 AGV Forklift

AGV Tugger (Image 1.3): Automatically guided tug vehicles pull non-motorized
load carriers behind them. They are used to transport heavy goods over longer
distances and may have several stops along predefined routes in a factory or
warehouse.

Image 1.3 AGV Tugger

AGV Các loại xe tự hành có mặt trên thị trường:

This robot is designed with support from the SLAM method, where an
environmental map is created, allowing the mobile robot to safely reach its destination.
The robot receives commands via WLAN, then executes driving tasks and closed-loop
5
control. The AGV uses algorithms for localization, route planning, and load handling.

All KUKA autonomous AGV systems are equipped with KUKA Navigation
Solution software, making them flexible and mobile to precisely meet various
requirements.

Image 1.4 KUKA AGV Robot

TPA SLAM AGV (Image 1.5): The SLAM-based AGV uses differential two-
wheel drive or stable steering wheels and is applied in environments with minor
changes and relatively good ground conditions. Environmental contours are used as
localization references, combined with odometry and laser-scan localization.

This enables high positioning accuracy without requiring infrastructure such as

reflectors or magnetic markers. It is a fully autonomous navigation method.

Advantages: reduced implementation difficulty, scalability to automatic obstacle

avoidance, and greatly improved flexibility. A compact AGV forklift using SLAM
navigation technology is an example.

Image 1.5 TPA SLAM AGV

6
 1.1.4 Applications of line-following delivery robots

The use of line-following delivery robots is to transport materials from one

place to another based on their weight, with each weight corresponding to a
predefined location. In industries, the requirement to transport products from one
production area to another often occurs across different buildings or even
separate blocks.
Conventional trucks or carts operated directly by humans are usually used,
which unintentionally reduces labor productivity and sometimes causes
inconveniences during transportation, such as not delivering according to
requirements, or even unfortunate accidents involving operators.
Therefore, automation in this area has emerged, using robots to follow a line
instead of installing rails, which are costly and inconvenient. The movement of
these robots depends entirely on the line, which is usually a drawn path or an
electrical wire as a magnetic concept.
Line-following delivery robots are often used to support automated production
processes, in military applications, for assisting humans, delivery services, and
especially in logistics and warehouse management.

1.2 Design objectives

Design a robot to load and deliver goods according to their weight, moving
along a line on the platform in Figure 1.6, with the following characteristics:
 Platform surface color: white
 Path color: black
 Path width: 26 mm
 Movement: in a plane
 Minimum speed: 0.1 m/s
3000
500 500 1500

Kết thúc
(1 kg)

R800 R800
1

R500
Kết thúc
(2 kg)
300
10 Bắt đầu

2000
Khu vực tải hàng
 Image 1.6 Platform demensions

The robot must meet the following operating requirements of the system:
 The robot moves from the “Start” position to the “Loading area” and stops.
 At the “Loading area,” a package weighing 1 kg or 2 kg is manually placed
on the robot. The robot can detect that the load has been placed.
 After receiving the load, the robot must move to the “End” position
corresponding to the package’s weight.
For the purpose of designing and manufacturing the line-following delivery
robot to track the platform, the following parameters must be considered: the
robot’s speed on the platform, its load capacity, its ability to change direction,
and the maximum tracking error when following the line.
Regarding speed, the robot must ensure successful delivery to the designated
location. Regarding load capacity, the robot must carry at least 2 kg. Regarding
direction changes, the robot must be able to follow curved paths with radii of 500
mm and 800 mm on the platform. Regarding maximum error, the deviation
during movement along straight or curved paths depends on the position
detection error from the sensors and the control system, and on the point selected
on the robot for determining line-tracking error

1.3 Design options for line-following delivery robots

Mechanical design:
 3-wheel type: 2 active wheels, 1 passive wheel
 4-wheel type: 2 active wheels, 2 passive wheels
 4-wheel type with independent drive
 4-wheel type using skid drive steering mechanism
Control structure:
 Centralized control
 Hierarchical control

11
Sensors:
 Line sensors:
 Camera
 Infrared sensor
 Photoresistor sensor
 Weight sensors:
 Load cell
 Proximity sensor
Controller:
 PID controller
 Lyapunov controller

12
CHAPTER 2 DESIGN SELECTION

2.1 Mechanical design

For a line-following car, the mechanical part acts as the frame, supporting
structure, and load-bearing component.
In addition, the mechanical part affects control and determines technical
characteristics after design such as: maximum speed, turning radius, and load
capacity.

2.1.1 Three-wheel type with two driving wheels and one passive wheel

Image 2.7 Xe dò line 2 chủ động – 1 bị động

Advantages:
 Three wheels are coplanar, in contact with the ground.
 Driving wheels at the back make it easy for the car to accelerate and increase
balance during movement.

Disadvantages:

13
  The front wheel tends to slip because the center of gravity is not on the axis
connecting the two motors, but this drawback can be overcome through the
controller.
In the design (figure 2.1), the car is driven by 2 rear wheels, each attached to
a separate motor. The 3rd wheel is a caster that can rotate in any direction.

Table 2.2 Line-following car specifications

Value Unit

Size 145 x 90 x mm
70

Average speed 1,17 m/s

Weight 0,21 kg

2.1.2 Four-wheel type with two independent driving wheels and two passive
wheels

In the design (figure 2.2), the car is driven by 2 rear wheels, with 2 caster
wheels in the front.

Image 2.8 Xe dò line 2 chủ động – 2 bị động

Advantages:

 Center of gravity in the middle helps the car take small-radius turns.

14
  With 4 wheels, the car is less likely to tip under heavy loads compared to 3-
wheel cars.

Disadvantages:

 Hard to ensure coplanarity for all 4 wheels.

 Easy to wobble when the center of gravity is far from the drive axle.

Table 2.3 Bảng thông số xe dò line

Value Unit

Size 182 x 132 mm

x 60
Average 1,07 m/s
speed
Weight 0,295 kg

2.1.3 Four independent driving wheels

Figure 2.3 design is driven by 4 wheels, each attached to an independent

motor.

Table 2.4 Line-following car specifications

Value Unit
Size 100 x 100 mm
x 40
Average 1,5 m/s
speed
Weight 0,2 kg

15
Image 2.9 Xe dò line Fireball

16
 2.1.4 Four-wheel type using skid drive steering

In the design (figure 2.4), the car is driven by 4 wheels, each motor controls
2 wheels on the same side.

Image 2.10 Xe dò line cơ cấu lái skid drive

Advantages:
 Can move on small-radius curves, slopes, center of gravity concentrated in
the middle.
 Load distributed evenly over all 4 wheels.

Disadvantages:
 Hard to control coplanarity with 4 wheels.
 Control algorithm is much more complex.
Table 2.5 Line-following car specifications

Value Unit
Size 175 x 153 x 40 mm
Average 2,5 m/s
speed
Maximum 4 m/s
speed
Weight 98 g

17
So sánh các thiết kế cơ khí xe dò line

Table 2.6 So sánh các kết cấu xe dò line

A B C D

Configuratio 4 wheels
2 driving 2 driving
n 4 independent using
wheels, 1 caster wheels, 2 caster
driving wheels skid
wheel wheels
drive

Flexible with Not flexible, Easier to

Turning Flexible with
large turning prone to control than
ability very small radius
radius slipping type C

Simple Simple
Control Simple because Complex because
because only 2 because only
complexity only 2 motors 4 motors
motors 2 motors

More
complex than
Structural More complex
type A, harder
complexity Simplest than type B Most complex
to keep
coplanar

Stability High, harder Relatively high, High, but

Low, easy to
when to tip because but prone to prone to
tip due to inertia
cornering more ground slipping due to slipping

contact points centrifugal force

2.2 Control structure

The electrical circuit of the line-following car consists of three main parts:
sensor circuit, control circuit, and motor driving circuit.
Two main methods to connect these parts are centralized control and
hierarchical control.

18
 2.2.1 Điều khiển tập trung

In centralized control (figure 2.5), a single microcontroller receives and

processes signals from sensors and motor encoders, runs the main program,
calculates control values, and transmits to the driver to control both motors.

Image 2.11 Cấu trúc điều khiển tập trung

2.2.2 Hierarchical control

In hierarchical control (figure 2.6), one microcontroller is used as master for

the main control program. Other slave microcontrollers handle separate tasks such
as: processing sensor signals, computing relative position of the car with respect
to the line and sending to the master; receiving encoder signals, computing motor
control laws to ensure motor operation according to master’s request, etc.
Signal exchange between MCUs may use standards like I2C, CAN, etc.
This reduces master workload and allows the robot to perform multiple tasks
simultaneously.

Image 2.12 Cấu trúc điều khiển phân cấp

19
 2.3 Types of sensors

2.3.1 Camera

Camera provides images containing useful information such as: surrounding

vision, obstacles, movement direction.
Key parameters: resolution and image processing chip speed. Higher
resolution = sharper image but higher cost and storage demand.

20
 Image 2.13 Camera Raspberry

Advantages: high accuracy.

Disadvantages: requires complex image processing algorithms, reducing
processing speed. Suitable only for slow mobile robots. High cost is also a
drawback.
2.3.2 Infrared sensor

Infrared sensor (figure 2.8): commonly used.

When IR beam from LED is reflected, the sensor generates voltage.
Advantages: very fast response (ns), simple, easy to use.
Disadvantages: easily affected by light.

Image 2.14 Cảm biến hồng ngoại TCRT5000

21
 2.3.3 Photoresistor

Image 2.15 CdS Photoresistor

Operates similarly to phototransistor: converts light energy into electrical

energy but only generates current (not voltage like phototransistor).
Need more than 2 sensors for accurate detection at curves.
Two sensor layouts: matrix and linear.

a) dạng ma trận b) dạng đường thẳng đơn lẻ

Image 2.16 Sensor layout

Matrix layout (Figure 2.10a): A method to determine the path. With many
sensors, more information is available to know the position and direction of the
car.
Disadvantage: uses too many sensors, so selection must consider size and
response speed of the car.
Linear layout (Figure 2.10b): Still provides enough information for
processing with fewer sensors than the matrix layout, so it is more suitable for line
detection.
Note: outermost sensors should not be too far or too close. If too far, accuracy is
lost in curves. If too close, line intersections are hard to detect.

Table 2.8 Comparison of sensor types

22
 Sensor type Camera Infrared sensor Phototransistor

Response time Slow (10–30 fps) Fast (~10 ms) Fast (≤1 ms)

Accuracy High Low Low

Cost High Low Low
Setup Complex Simple Simple

2.3.4 Load cell

Load cell is made of two parts: the strain gauge and the load element.

Image 2.17 Cấu tạo của Strain Gauge

Load cell works on Wheatstone bridge balance. When a force/load acts on the load cell, it

causes deformation (stretch or compress), which changes the length and cross-section of strain

gauge wires → resistance changes → output voltage changes proportionally to applied force.

23
 Image 2.18 Load cell structure

2.4 Controllers

Many controllers can be used for line-following racing cars: PD, PID, fuzzy, etc.
To apply these, a gyro sensor is usually required to measure acceleration state.

2.4.1 Control algorithm

Requirements: robot must follow straight and curved tracks, while ensuring high
speed and stability.
→ The group chose PID controller, because it is widely used for path tracking, can
handle errors and achieve required velocity.
Additionally, the team plans to use PID + path memory. Though self-learning
algorithm is complex, combining sensors to memorize the path allows the robot to store
acceleration and angle states → enabling faster and more stable movement.
PWM
Error
+ PID 1 Driver 1
_
Encoder 1 Motor
Desired value + Error Flowing tracking Actual value
_

Error PWM
+ PID 2 Driver 2
_ Motor
Encoder 2

Sensor

24
 Image 2.19 Control system block diagram

2.4.2 Comparison algorithm

Uses ON/OFF sensor signals to determine the relative position between the robot
and the line.
Advantages:
 Simple and fast.
 Accuracy depends on signal processing and sensor spacing.

Image 2.20 Comparison algorithm

Summary:

Objective: Design a line-following racing car to run as fast as possible while

meeting accuracy requirements.
Track specifications:
 Line color: black
 Surface color: white
 Line width: 26 mm
 Surface: flat
Table 2.9 Robot specifications

Specification Value
Minimum turning radius ρmin = 500 mm
Maximum line tracking error emax = ±3 mm
25
 Expected velocity vmax = 0.7 m/s
Minimum load capacity mmin = 2 kg

Table 2.10 Preliminary selection

Three-wheel type with two

Mechanical design driving wheels and one
caster wheel

Driving
wheel

Caster
wheel

Control structure Centralized control

Microcontro
ller

Driver

Sensors Line sensor (TCRT5000)

and load cell

Line sensor
(TCRT5000)

26
 Load sensor
(Loadcell)

PID controller for motors,

Controller Lyapunov stabilization,
comparison algorithm for line
detection

27