MS SYNOPSIS
Graph Coloring Algorithm in AI-Driven Task Scheduling
Submitted by
Unzila UW-24-MATH-MS-0l6
Atiqa UW-24-MATH-MS-0l7
Rabia UW-24-MATH-MS-032
Qurat-ul-AIN UW-24-MATH-MS-021
Noshaba UW-24-MATH-MS-010
Supervisor
Mr. Zamir Hussain
Co-Supervisor
Terence Tao
Graph Coloring Algorithm in AI-Driven Task Scheduling
Serial No. Table of Contents Page No.
01 Abstract 3
02 Introduction 3
03 Research Problem 4
04 Research Objectives 5
05 Background 5
06 Methodology 6
07 Resource Required 9
08 Project Plan 10
09 References 12
10 Undertaking 13
Page 2 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
01. Abstract
This paper introduces a new framework for AI-based task scheduling that makes use of
mixed graphs' optimal coloring. In a mixed graph structure, tasks with integer durations,
release times, and due dates are represented as vertices and edges, along with constraints
on precedence and simultaneous execution. We establish a one-to-one correspondence
between valid colorings and schedules that respect both undirected (conflict) and directed
(precedence) edges. Assigning the smallest possible set of time slots (colors) under these
constraints reframes the scheduling goals of minimizing makespan or maximum lateness.
Because of this equivalence, AI systems can use sophisticated graph coloring algorithms
to solve dynamic scheduling problems involving multiple processors. By extending
classical scheduling theory through a graph-theoretic lens and supporting non-
preemptive, interdependent operations, the method makes intelligent task management
possible through effective, rule-consistent orchestration.
02. Introduction
AI-based systems more and more depend on advanced scheduling techniques to schedule
interdependent, resource-scarce, and time-critical tasks. With autonomous robot
coordination, real-time decision-making in multi-agent systems, scheduling under
challenging constraints is essential. Classic scheduling models usually fail in addressing
dynamic priorities, concurrent need for execution, and multi-processor task dependencies.
This work tackles the above challenges by reducing AI task scheduling to a mixed graph
colouring problem to facilitate efficient and rule-based management of tasks with integer
durations, release dates, and due dates.
A mixed graph contains both directed arcs and undirected edges so that it can model both
precedence constraints and mutual exclusions simultaneously. It is a strong abstraction
for representing scheduling problems where some jobs need to be done before others,
while others cannot be executed concurrently. Past studies have demonstrated that job-
shop and general-shop scheduling problems with unit or integer-time tasks are actually
Page 3 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
equivalent to mixed graph colouring problems in which the colour of each vertex is a
time slot, and valid colourings observe both conflict and sequencing rules .[1]
In our framework, we generalize this equivalence to AI scheduling scenarios where jobs
can be interruptible, need to be executed simultaneously, or have multiple processors. We
establish that finding a feasible and optimal schedule to minimize traditional objectives
like makespan (Cₘₐₓ) or maximum lateness (Lₘₐₓ) is computationally equivalent to finding
an optimal colouring of the respective mixed graph. This graph-theoretic model makes it
possible to apply known colouring algorithms to a new type of AI scheduling problems,
opening a gap between combinatorial optimisation and intelligent task scheduling.
The value added by this work is its integrating framework that links AI scheduling with
the algorithmic richness and theoretical strength of graph colouring. This correspondence
not only supplies a solid analytical foundation but also unleashes the potential for
scalable, general-purpose scheduling algorithms that can be applied to a broad variety of
AI applications ranging from smart manufacturing to real-time robotics.
03. Research Problem
AI-based systems need effective task scheduling to manage intricate dependencies and
limited resources. Graph coloring algorithms provide a promising framework by
representing tasks as graph nodes and conflicts as edges, giving "colors" to avoid overlap.
Conventional coloring techniques fail in dynamic settings with hard real-time
requirements. Our research has two goals: to create an adaptive graph coloring algorithm
specific to AI-based task scheduling, and to implement the algorithm with performance
and scalability in mind. The objective is to reduce execution time, prevent resource
conflicts, and maximize throughput. Through the combination of AI methodologies and
graph theory, the suggested model will overcome the shortcomings of current methods
and enhance scheduling in multi-agent and cloud-based AI systems.
04. Research Objectives
• To represent AI-oriented task scheduling issues as graph coloring problems, modeling
tasks, dependencies, and resource conflicts accurately.
Page 4 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
• To create an adaptive graph coloring algorithm that is sensitive to dynamic task
priorities and real-time constraints.
• To incorporate AI methods (e.g., reinforcement learning or heuristics) into graph
coloring for better scheduling choices.
• To assess the efficiency of the suggested algorithm in terms of execution time,
resource utilization, and system performance based on simulated or real artificial
intelligence workloads.
05. Background
Scheduling tasks forms the backbone of AI-driven system operation, where tasks are
frequently interdependent, have shared resources in finite quantities, and require time-
critical execution. Ranging from robotic tasks and real-time planning in multi-agent
settings to production lines in intelligent manufacturing, effective scheduling decides
system responsiveness and system throughput. Classical scheduling formulations like
job-shop or flow-shop models usually consider fixed constraints and linear precedence
relationships. These models are inadequate when tasks must run concurrently on multiple
processors or when both concurrency and task order have to be preserved in real time.
For such problems, researchers have developed graph-theoretic methods that enable
scheduling problems to be formulated using structural abstractions. Of these, mixed
graphs have been very useful. A mixed graph includes a set of vertices (activities), a set
of directed arcs (precedence relations), and a set of undirected edges (resource or conflict
constraints). Colouring vertices of such graphs under clearly specified rules allows
converting scheduling problems into graph colouring problems. In a correct colouring of
a mixed graphG=(V,A,E), directed arcs (vi, vj) impose that the colour of vi is less than or
equal to that of vj (to maintain precedence), while undirected edges [vi, vj] imply vi and
vj must have distinct colours (to avoid concurrent execution on common resources). The
aim is to discover the chromatic number, the smallest number of colours to meet all
constraints—analogous to minimizing makespan or maximal lateness in scheduling.
Sotskov and Mihova played a central role in formalizing this correspondence. Their
works of 2021 and 2024 demonstrated that numerous classical and generalised
Page 5 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
scheduling problems—including unit-time job-shop problems, intricate multi-processor
obs with interruptibility and simultaneityy—can be formal translated into mixed graph
colouring problems. In particular, their latest work presented a graph-based approach to
accommodating simultaneity by representing such requirements as cycles within the
mixed graph. This allowed for scheduling tasks that need to execute concurrently
between processors while retaining integer-based start and due dates [2].
Even with these developments in graph-based scheduling theory, their real
applications in AI task management have not been explored. Task sets in AI contexts
tend to be dynamic, consist of interruptible operations, and need to be optimized in
real time under varying constraints. Existing scheduling approaches either depend on
heuristics with no formal guarantees or fail to scale with complexity. There is an
evident requirement for leveraging the theoretical power of mixed graph colouring
into a computationally effective framework that is appropriate for intelligent, rule-
based AI scheduling.
This research builds upon the theoretical groundwork of graph colouring in scheduling
and aims to create a practical, AI-compatible framework that uses mixed graph
colouring to produce optimized, conflict-free, and scalable schedules for multi-
processor AI tasks with precedence, simultaneity, and timing constraints.
06. Methodology
This research develops a graph-theoretic methodology to solve AI-driven task
scheduling problems by reducing them to optimal colouring problems on mixed
graphs. The approach is motivated by the formal equivalence established by Sotskov
and Mihova (2021, 2024), who demonstrated that scheduling problems involving
precedence, simultaneity, and mutual exclusion can be faithfully represented and
solved using mixed graph colouring.
1. Problem Representation via Mixed Graphs
We define a mixed graph G= (V, A ,E), where:
1. 𝑉= {𝑣1 ,𝑣2,...,𝑣𝑛} is the set of tasks (vertices),
Page 6 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
2. A⊆V×V is the set of directed arcs modeling precedence constraints,
3. E ⊆ { { 𝑣𝑖,𝑣𝑗} :𝑣𝑖,𝑣𝑗∈𝑉,𝑖≠𝑗} is the set of undirected edges modeling
mutual exclusion (e.g., resource conflicts).
In cases where a group of tasks must be performed simultaneously, we encode this
by introducing a cycle (i.e., closed directed path) over the involved vertices.
2. Mathematical Framework: Mixed Graph Colouring
Let c:V→N be a colouring function assigning time slots (positive integers) to tasks.
A valid colouring satisfies:
1. Precedence constraint: For every arc (𝑣𝑖,𝑣𝑗)∈𝐴
𝑐(𝑣𝑖)≤𝑐 (𝑣j )
2. Conflict constraint: For every edge {𝑣𝑖 ,𝑣𝑗}∈𝐸,
𝑐(𝑣𝑖)≠𝑐(𝑣𝑗)
3. Simultaneity constraint: For any cycle 𝐶={𝑣𝑖1,𝑣𝑖2,...,𝑣𝑖𝑘,𝑣𝑖1}⊂𝐴
C {𝑣𝑖1)=𝑐(𝑣𝑖2)=...=𝑐(𝑣𝑖𝑘)
The objective is to minimize the chromatic number of the mixed graph χ(G), defined
as:
𝜒(𝐺)=minmax𝑐(𝑣𝑖)
This corresponds to minimizing the makespan 𝐶max of the schedule.
3. Algorithm Development and Optimization
We propose a modular framework:
Graph construction module: Encodes AI scheduling instances into mixed graphs
using task metadata (precedence, resources, simultaneity).
Colouring engine: Implements or adapts algorithms for mixed graph colouring, such
as:
∎Greedy colouring with constraint propagation:
Page 7 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
The greedy algorithm for graph colouring—one of the simplest and earliest colouring
techniques—was introduced in the 1950s–1960s as part of foundational work in graph
theory and combinatorics. The first formal reference to a greedy (or sequential) colouring
algorithm is often credited to David Welsh and Michael Powell (1967).
∎Integer programming for small graphs:
Integer programming (IP), specifically Integer Linear Programming (ILP), is a
powerful method for finding optimal solutions when dealing with small or moderately
sized task graphs.l/
∎Heuristics (e.g., local search, simulated annealing) for scalability:
Heuristics are approximate algorithms designed to find good-enough solutions
quickly—especially when exact algorithms (like integer programming) are too
slow or infeasible due to problem size. In the context of AI-driven task scheduling
using graph colouring, heuristics help manage large, complex task graphs where
optimal scheduling is computationally hard (NP-hard).
Simultaneity handler: Ensures cycle-based encoding for groups of simultaneous tasks.
Validation layer: Translates coloured graphs back into executable schedules.
Simulation and Computational Approach
Graph coloring algorithms in AI-driven task scheduling, several computational software
tools can be effectively used depending on the goals (modeling, simulation,
optimization, or implementation).
Python:
Use: Algorithm development, AI integration
Libraries like NetworkX (for graph modeling), PuLP/CVXPy (for optimization),
and TensorFlow/PyTorch (for AI).
MATLAB:
Use: Mathematical modeling and simulations.
Strong support for matrix-based graph representations and scheduling simulations.
Page 8 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
SageMath:
Use: Graph theory, symbolic math
Built-in graph coloring functions, integrates Python.
Comparative Analysis
This research conducts a comparative evaluation of various computational strategies for
solving the AI-driven task scheduling problem modeled through mixed graph colouring.
The goal is to assess the efficiency, scalability, and solution quality of both exact and
heuristic approaches in diverse scheduling scenarios characterized by precedence,
conflict, and simultaneity constraints.
To reflect the dual focus on theoretical rigour and computational feasibility, we compare
three main algorithmic classes:
1. Exact Methods (Baseline – Integer Linear Programming)
2. Greedy Colouring with Constraint Propagation
3. Metaheuristic Approaches (Simulated Annealing, Local Search)
While the core application is AI-driven task scheduling, the comparative insights from
this study extend to other domains where constraint-driven optimization is required, such
as,Distributed AI planning,Smart manufacturing workflows,Parallel resource
allocation,Robotics and multi-agent systems.
07. Resource Required
To conduct the simulation and analysis of graph colouring algorithms in AI-driven
task scheduling, the project requires access to both software tools and moderate
computational hardware.
Data Resources
Task scheduling graphs will be synthetically generated using custom Python scripts.
These graphs will model real-world AI environments with precedence constraints, mutual
Page 9 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
exclusions, and simultaneous task execution. No external datasets are needed, ensuring
controlled experimental design.
Software Tools
Python 3.x: Core development environment.
NetworkX: For representing mixed graphs and managing task dependencies.
PuLP + CBC/Gurobi: For solving integer programming formulations of scheduling
problems.
Custom heuristic modules: Implementing greedy colouring, local search, and simulated
annealing.
Matplotlib/Seaborn: For result visualization and comparative analysis.
Hardware Requirements
Simulations will run on a personal computer or standard lab workstation with:
At least 8 GB RAM, 4-core CPU, and SSD storage.
For large-scale scenarios, optional access to cloud platforms (e.g., Google Colab, Azure
Notebooks) may be used.
All required resources are either open-source or available via academic/free licenses,
ensuring accessibility throughout the research.
08. Project Plane
Phase 1: Literature Review & Theoretical Research (Months 1-3)
Review foundational concepts in scheduling theory, graph colouring, constraint
satisfaction, and AI task orchestration.
Study key works by Sotskov & Mihova (2021, 2024) and related papers on greedy,
heuristic, and ILP-based scheduling.
Formulate the problem mathematically using mixed graphs and define key constraints
(precedence, conflict, simultaneity).
Page 10 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
Phase 2: Mathematical Modelling & Algorithm Development (Months 4-6).
Develop the greedy colouring algorithm with constraint propagation.
Implement heuristic techniques including local search and simulated annealing.
Set up ILP models for exact scheduling using PuLP and solvers (e.g., CBC, Gurobi).
Prepare synthetic task datasets for evaluation.
Phase 3: Simulation & Computational Analysis (Months 7-9)
Run experiments across various task graph sizes (50–1000 nodes).
Evaluate algorithms on feasibility, makespan, lateness, and runtime.
Perform comparative analysis between ILP, greedy, and metaheuristic methods.
Document simulation results and generate visualizations (e.g., Gantt charts, colouring
maps).
Phase 4: Comparative Study & Optimization (Months 10-11)
Draft the thesis or research paper, including all core sections (abstract, intro,
methodology, results, etc.).
Phase 5: Documentation & Thesis Writing (Months 12-12.5)
• Draft research paper, thesis, and final report.
• The findings should be sent to reviewers for assessment and publication in conferences
Final Phase: Review, Presentation & Submission (Last Two Weeks)
• The researcher must present their final dissertation and project materials for
evaluation to the research committee.
• The organized scheduling mechanism maintains regular progression while researchers
identify and optimize mathematical equations and algorithm code developments.
Page 11 of 12
Graph Coloring Algorithm in AI-Driven Task Scheduling
This research will be conducted over one year, divided into distinct phases to ensure
systematic development, implementation, and evaluation of quantum computing
frameworks. The timeline includes literature review, mathematical modelling, algorithm
development, simulation, comparative analysis, and documentation.
09. References
1. Sotskov, Y.N. & Mihova, E.I. (2021). Scheduling multiprocessor tasks with equal
processing times as a mixed graph coloring problem. Algorithms, 14(8), 246.
2. Mihova, E.I. & Sotskov, Y.N. (2024). Mixed Graph Colouring as Scheduling a
Partially Ordered Set of Interruptible Multi-Processor Tasks with Integer Due
Dates. Algorithms, 17(7), 299.
10. Undertaking
We, Unzila, Atiqa, Rabia, Qurat-ul-AIN, and Noshaba, students of MS Mathematics,
University of Wah, Wah Cantt, declare that our research proposal titled "Graph
Colouring Algorithm in AI-Driven Task Scheduling" is original and based entirely on our
work, except where proper references have been provided.
______________________ ______________________
(Signature of Students) (Signature of Supervisor)
Page 12 of 12
