SIMON THOMPSON

IROS Conference 2025 Conference Paper

Multi-PrefDrive: Optimizing Large Language Models for Autonomous Driving Through Multi-Preference Tuning

• Yun Li
• Ehsan Javanmardi
• Simon Thompson
• Kai Katsumata
• Alex Orsholits
• Manabu Tsukada

This paper introduces Multi-PrefDrive, a framework that significantly enhances LLM-based autonomous driving through multidimensional preference tuning. Aligning LLMs with human driving preferences is crucial yet challenging, as driving scenarios involve complex decisions where multiple incorrect actions can correspond to a single correct choice. Traditional binary preference tuning fails to capture this complexity. Our approach pairs each chosen action with multiple rejected alternatives, better reflecting real-world driving decisions. By implementing the Plackett-Luce preference model, we enable nuanced ranking of actions across the spectrum of possible errors. Experiments in the CARLA simulator demonstrate that our algorithm achieves an 11. 0% improvement in overall score and an 83. 6% reduction in infrastructure collisions, while showing perfect compliance with traffic signals in certain environments. Comparative analysis against DPO and its variants reveals that Multi-PrefDrive’s superior discrimination between chosen and rejected actions, which achieving a margin value of 25, and such ability has been directly translates to enhanced driving performance. We implement memory-efficient techniques including LoRA and 4-bit quantization to enable deployment on consumer-grade hardware and will open-source our training code and multi-rejected dataset to advance research in LLM-based autonomous driving systems. Project Page (https://liyun0607.github.io/).

AAMAS Conference 2007 Conference Paper

Intelligent Agent Framework for Order Entry and Management

• Thuc Duong Nguyen
• SIMON THOMPSON

This paper describes an agent system we have built to handle order entry and management issue in business computing and specialized in the telecommunication domain. Our system is implemented using only industry approved standards and it can both manage complex workflows as well as handle user context dependant service selection. Furthermore, it has the abstract goal representation mechanism, which in turn can detect and reflect to changes in its working environment, leading to more accurate capturing of user's requirements. We also detail a number of prototype applications that has been built using our technology and highlight how our approach addresses the business issues around each of these applications.

AAMAS Conference 2006 Conference Paper

CONOISE-G: Agent-based Virtual Organisations

• W. T. Luke Teacy
• Nicholas R. Jennings
• Michael Luck
• Stuart Chalmers
• Nir Oren
• Timothy J. Norman
• Alun Preece
• Peter M. D. Gray

KER Journal 2002 Journal Article

RAMASD: a semi-automatic method for designing agent organisations

• Anthony Karageorgos
• Nikolay Mehandjiev
• SIMON THOMPSON

Designing realistic multi-agent systems is a complex process, which involves specifying not only the functionality of individual agents, but also the authority relationships and lines of communication existing among them. In other words, designing a multi-agent system refers to designing an agent organisation. Existing methodologies follow a wide variety of approaches to designing agent organisations, but they do not provide adequate support for the decisions involved in moving from analysis to design. Instead, they require designers to make ad hoc design decisions while working at a low level of abstraction. We have developed RAMASD (Role Algebraic Multi-Agent System Design), a method for semi-automatic design of agent organisations based on the concept of role models as first-class design constructs. Role models represent agent behaviour, and the design of the agent system is done by systematically allocating roles to agents. The core of this method is a formal model of basic relations between roles, which we call role algebra. The semantics of this role-relationships model are formally defined using a two-sorted algebra. In this paper, we review existing agent system design methodologies to highlight areas where further work is required, describe how our method can address some of the outstanding issues and demonstrate its application to a case study involving telephone repair service teams.