| On this page, I have summarised my dissertation project ideas. Many of the projects focus on Agent-Based Modelling and Simulation, which is my primary research area. More recently, I have been exploring how Generative AI can be integrated into the Agent-Based Modelling and Simulation lifecycle. If you would like to learn more, please visit my ABM Blog. |
| Projects with a strikethrough title are currently not available. |
| Generative AI Related Dissertation Projects |
| Advanced Prompt Design and Quality Assessment for Agent Decision-Making in Agent-Based Social Simulation |
|
Social simulation employs agent-based models, in which autonomous agents interact within defined environments according to behavioural rules, allowing complex collective patterns to emerge from simple individual decisions. This research project is related to one of my recent blog posts, "The Generative ABM Experiment (1/3): From Concept to First Prototype." In it, I discussed the question of how to integrate LLMs into social simulations in order to enable agents to mimic human-like decision-making behaviour. While I succeeded in getting something running that produces correct patterns at the system level, I realised that the communication between agents and LLMs often produced non-matching responses (e.g. LLM providing "no" as a decision while the related LLM reasoning suggests it should be a "yes"). Related Blog Post 'The Generative ABM Experiment (1/3)' The aim of this project is to improve the robustness of the prompts to deliver logically correct outcomes. This includes evaluating the current situation and developing and testing novel prompt engineering strategies. Such strategies could include in-prompt validation of responses, considering different levels of memory required by agents to create realistic responses and perhaps learn from previous decisions. There might also be technical issues related to the communication between the Python frontend and the LLM backend, leading to the non-matching decision/reasoning pairs, which should be investigated. Keywords: Artificial Intelligence; Games; Machine Learning; Simulation Methodologies: Algorithms / Architecture; Programming / Implementation |
| RAG Systems for Enhancing Agent-Based Modelling |
|
Standard LLMs often struggle with the specific technical requirements of Agent-Based Modelling (ABM), such as specialised implementation logic or nuanced decision-making frameworks. Retrieval-Augmented Generation (RAG) offers a way to ground LLMs in domain-specific knowledge to bridge this gap. The aim of this project is to design and implement a RAG system tailored for ABM tasks. You will explore how "ragging" can improve the quality of LLM outputs across the design, decision-making, and implementation phases of simulation development. Keywords: Artificial Intelligence; Big Data; Machine Learning; Simulation; Generative AI Methodologies: Algorithms / Architecture; Programming / Implementation; Data Analysis |
| Interactive Interface for the RAT-RS Reporting Standard |
|
The Rigour and Transparency Reporting Standard (RAT-RS) was developed to improve how data use is documented in agent-based modelling. Social simulation employs agent-based models, in which autonomous agents interact within defined environments according to behavioural rules, allowing complex collective patterns to emerge from simple individual decisions. To increase the adoption of the RAT-RS, researchers need an interactive and accessible tool to navigate these standards. Information about RAT-RS (Journal Paper) The aim of this project is to implement an interactive LLM-supported Streamlit application and searchable database for the RAT-RS framework. Following previous work (see blog post links below), you will develop LLM-driven question/answer sets to guide researchers through the documentation process and evaluate the quality of the resulting responses using quality assessment methods such as semantic analysis and similar methods. LLM Support for Creating RAT-RS Reports: Blog Post (1/2) LLM Support for Creating RAT-RS Reports: Blog Post (2/2) Keywords: Artificial Intelligence; Human-Computer Interaction; Simulation; User Interfaces; Generative AI Methodologies: Programming / Implementation |
| Streamlit App for LLM-Driven Creation of Agent-Based Models |
|
Social simulation employs agent-based models, in which autonomous agents interact within defined environments according to behavioural rules, allowing complex collective patterns to emerge from simple individual decisions. The Engineering Agent-Based Social Simulation (EABSS) framework provides rigorous structure for conceptual model design but suffers from a steep learning curve and demanding moderator requirements. While the recently published conversational AI script demonstrates that LLMs can generate complete ABSS model specifications, its current implementation is linear and passive, users merely submit prompts and observe. This project addresses the critical gap between automated generation and genuine human-AI collaboration, transforming the script into an interactive design environment where modellers actively steer model development through meaningful decision points. JASSS Journal Paper with Related Script The aim of this project is to design, implement, and evaluate an interactive Streamlit application that transforms the EABSS prompt script into a conversational, roleplay-inspired tool. The application will enable modellers to make consequential design decisions at each EABSS step, generating multiple branching pathways and diverse model specifications. Keywords: Artificial Intelligence; Games; Simulation; User Interface Methodologies: Programming / Implementation; Data Analysis |
| Exploring the Potential of Using Generative AI for Test-Driven Simulation Modelling |
|
Test-Driven Development (TDD) is a software development approach where tests are written before the actual code, guiding development and ensuring functionality through iterative testing and refinement. Whilst TDD has proven effective in traditional software engineering contexts, its application to Simulation for Decision Support remains relatively unexplored, despite simulation model development sharing fundamental characteristics with conventional software projects. Generative AI technologies, particularly LLMs capable of code generation and understanding complex domain requirements, present unprecedented opportunities to automate and accelerate the Test-Driven Simulation Modelling (TDSM) workflow by intelligently generating both test cases and corresponding simulation code that satisfies those tests. The aim of this project is to explore how Generative AI can be leveraged within TDSM to automatically generate test cases and the corresponding simulation code required to pass those tests. You will investigate the capabilities and limitations of contemporary Generative AI systems in understanding simulation requirements, producing meaningful test specifications, and generating valid implementations that adhere to TDD principles. Information about DES Information about TDSM Keywords: Software Engineering Methodology: Programming / Implementation |
| Back to Top |
| Decision Support Dissertation Projects |
| A Simulation Testbed for Digital Twins |
|
Digital Twins (virtual representations of physical systems that are continuously updated with real-world data to mirror their behaviour and enable predictive analysis) are increasingly used across application areas such as healthcare, operations research, and sustainability. Yet we lack environments to test their interactions and performance under varying conditions. The aim of this project is to create different types of Digital Twins and a simulation testbed for evaluating these. This platform will allow you to demonstrate how Digital Twins can support real-time decision-making in complex environments. Keywords: Artificial Intelligence; Simulation; Optimisation Methodologies: Algorithms / Architecture; Programming / Implementation |
| Simulating Smart Grids and Market Dynamics in National Electricity Networks |
|
Modern electricity networks are transitioning from centralised systems to complex, decentralised ecosystems. This shift introduces significant volatility that traditional linear models fail to predict. Agent-Based Modelling (ABM) is a computational framework used to simulate the actions and interactions of individual and collective autonomous agents to understand the behaviour of a system and what governs its outcomes. By utilising agent-based modelling, we can simulate the "emergent behaviour" of a grid where thousands of autonomous actors, from households with solar panels to industrial providers, interact. This project provides a vital tool for stress-testing the resilience of national infrastructure against shifting market regulations and fluctuating consumer demand without risking real-world failure. The aim of this project is to develop and implement a robust ABM in Python to simulate the dynamics of a national smart grid. You will design intelligent agents representing various stakeholders, including regulators, network operators, prosumers, and policy makers, to investigate "what-if" scenarios regarding regulatory changes and technology adoption, ultimately assessing their impact on grid stability and market competitiveness. Keywords: Artificial Intelligence; Optimisation; Simulation Methodologies: Programming / Implementation; Data Analysis |
| Agent-Based Simulation of Worker Exploitation in the Gig Economy |
|
The gig economy's rapid expansion has created unprecedented flexibility for workers whilst simultaneously raising concerns about exploitation through precarious employment, algorithmic management, and power imbalances. Understanding these complex socio-economic dynamics requires sophisticated modelling approaches that can capture emergent behaviours and interactions between workers, platforms, and regulatory frameworks. Agent-based social simulation offers a powerful methodology for investigating how exploitation manifests through individual decisions and systemic structures, enabling evidence-based policy recommendations and deeper insights into worker vulnerability patterns within platform-mediated labour markets. The aim of this project is to develop a comprehensive agent-based simulation model to investigate worker exploitation dynamics in gig economy platforms. The simulation will implement reusable, extensible components using Python (AgentPy or Mesa) or a specialised platform (Gama), conducting a complete simulation study lifecycle to analyse exploitation mechanisms and evaluate potential interventions. Information about Worker Exploitation Keywords: Artificial Intelligence; Simulation; Statistics Methodology: Algorithms / Architecture; Programming / Implementation |
| Back to Top |
| Miscellaneous Dissertation Projects |
| Student-led Project |
In general, I am happy to consider projects which relate in some form to the two topics mentioned below. If you have an idea which you believe aligns with those, please email me a brief description of your idea (up to 150 words), including project motivation and project aim.
Methodology: Algorithms / Architecture; Programming / Implementation; Training ML Models |
| Serious Game Development for Teaching Systems Thinking |
|
Systems thinking is a holistic approach to problem-solving that considers how parts of a system interact, influence each other, and contribute to the overall behaviour of the whole. Traditional educational methods often struggle to convey the dynamic, interconnected nature of complex systems, relying on static diagrams and theoretical explanations that fail to capture emergent behaviours and feedback loops. Serious games offer an interactive medium through which learners can experiment with system parameters, observe consequences of their decisions in real-time, and develop intuitive understanding of systemic relationships that would be difficult to grasp through conventional pedagogical approaches. The aim of this project is to develop a 2D simulation game specifically designed for teaching systems thinking principles. You will either adapt and clone existing systems thinking games or design an original game concept of your choice, creating an engaging platform that enables players to explore complex system dynamics, feedback mechanisms, and the non-linear relationships that characterise real-world systems across various domains. Information about 'systems thinking' Example of such a serious gane Keywords: Artificial Intelligence; Games; Simulation Methodology: Algorithms / Architecture; Programming / Implementation |
| Reimplementing the "Little Computer People" Game Using Modern AI |
|
"Little Computer People," released in 1985, represents one of the first mainstream agent-based social simulation games, pioneering the concept of autonomous virtual characters living within a digital environment and responding to player interactions with emergent behaviours. The game's groundbreaking approach to simulating believable artificial life laid foundational concepts for subsequent simulation genres, yet its AI capabilities were constrained by the computational limitations of its era. The aim of this project is to reimplement the "Little Computer People" concept using contemporary AI approaches to create an enhanced agent-based social simulation. You will develop autonomous virtual characters capable of sophisticated behaviours through modern pathfinding algorithms and decision-making systems, exploring how current AI technologies can breathe new life into this pioneering game concept whilst maintaining its charm and expanding its interactive possibilities. Information about the original game Keywords: Artificial Intelligence; Games; Machine Learning; Optimisation; Simulation Methodology: Algorithms / Architecture; Programming / Implementation |
| Developing an AI-Driven Card Game for Energy Awareness amongst Elderly Users |
|
The intersection of gamification and behavioural change has demonstrated significant potential in promoting sustainable practices, particularly in energy consumption. Elderly populations face unique challenges in engaging with abstract concepts such as energy usage, often struggling with traditional awareness campaigns that lack tangible feedback mechanisms. Card games, with their familiar mechanics and social engagement potential, offer an accessible medium for addressing these challenges whilst providing cognitive stimulation. Integrating artificial intelligence-driven opponents creates opportunities for adaptive difficulty, personalised feedback, and sustained engagement without requiring human co-players, addressing the social isolation many elderly individuals experience whilst simultaneously promoting environmental consciousness through gameplay mechanics that mirror real-world energy decisions. The aim of this project is to develop a Yaniv card game featuring AI-driven opponents specifically designed to enhance energy consumption awareness amongst elderly users. For the implementation you can use a programming language or game engine of your choice. Information about Yaniv Keywords: Artificial Intelligence; Games; Simulation Methodology: Algorithms / Architecture; Programming / Implementation |
| Energy-Efficient Data Forwarding in Mobile Wireless Sensor Networks |
|
Mobile Wireless Sensor Networks (MWSNs) deployed on wildlife present unique energy management challenges, as replacing batteries on free-roaming animals is impractical and disruptive. Unlike traditional static sensor networks, wildlife movement creates dynamic connectivity issues where network paths constantly reconfigure as animals traverse their habitats. Sensors must balance capturing behavioural data, maintaining communication links, and preserving battery life for extended deployments. Optimising data forwarding in such environments requires understanding both animal movement patterns and network computational constraints, creating a complex multi-agent system where behavioural agents (animals) and software agents (sensors) interact within an optimisation framework aimed at minimising energy expenditure whilst maintaining transmission reliability. The aim of this project is to develop a simulation model that evaluates energy-efficient data forwarding strategies in mobile wireless sensor networks deployed on wild elephants. You will model a hybrid system combining behavioural agents representing elephant movement patterns, software agents simulating sensor nodes and their communication protocols, and optimisation algorithms for determining efficient network routing strategies. For this project you can use either an existing simulation platform (e.g. AnyLogic or Gama) or build your simulation from scratch using a Python library (e.g. AgentPy or Mesa). Keywords: Artificial Intelligence;Simulation;Optimisation Methodology: Programming / Implementation; Algorithms / Architecture |
| Creating AI-Driven AnyLogic Extensions in Python |
|
...
Keywords: ... Methodology: ... |
| Back to Top |
| Advice for Dissertation Students |
| Please check out the Teaching Page for to find the sought-after advice. |
| Back to Top |