Evolution of AI SystemsFocused on the development of network topologies under incomplete information scenarios and computational complexity of formal grammars. Utilising NEAT (Neural Evolution of Augmented Topologies) and hyper-NEAT algorithms, the project aims to create an open-source Julia package to facilitate future research.
This work combines evolutionary theory, AI coordination, and reinforcement learning to better understand the behaviour of AI systems developed using evolutionary algorithms. Special interest in pathologies of such approaches and alignment issues with developing AI systems in this way.
Contact Dr Matishalin Patel, Patel@hull.ac.uk for more details.
Net-Zero Emissions Aviation: Developing Hydrogen Energy Infrastructure at Airports
Decarbonising aviation is essential for the UK's net zero emissions target by 2050, with plans for all airport operations and domestic flights in England to reach net zero by 2040. Hydrogen is a promising aviation fuel, but significant advancements in airport infrastructure are needed. While some airports, like Bristol, are exploring hydrogen infrastructure, technical knowledge gaps persist. Current research lacks detailed designs to integrate hydrogen with existing energy systems and meet net zero flight requirements. This project will evaluate, design, and optimise hydrogen-based airport infrastructure, covering production, storage, and transportation to support domestic hydrogen aircraft. Using a bottom-up approach, we will model hydrogen demand based on flight operations and assess devices' efficiency and risks. Case studies with a major UK airport and a hydrogen aircraft prototype will validate findings. Our research aims to fill critical gaps, providing actionable insights for sustainable aviation infrastructure and supporting the UK's 2040 targets.
For more information: Net-Zero Emissions Aviation: Developing Hydrogen Energy Infrastructure at Airports.
Multi-agent Large Language Model
Dr. Zekun Guo is leading a focused research group within DAIM to explore cutting-edge applications of Large Language Models (LLMs), particularly in Retrieval-Augmented Generation (RAG), agent systems, and multi-agent frameworks. The team includes Eva Sousa, Dr. Julius Mboli, Dr. Kenneth Wertheim, Dr. Matishalin Patel, and Dr. Temitayo Matthew Fagbola, bringing together a wealth of expertise and diverse perspectives.
The group aims to develop and evaluate multi-agent systems capable of addressing real-world challenges, such as decision-making in hospitals, factories, or power systems. These systems will leverage LLMs to generate and coordinate decisions, providing a robust platform to test their efficacy in complex environments. By combining the team’s innovative ideas and technical skills, this initiative seeks to establish a structured pathway for impactful research.
Dr. Guo is currently exploring frameworks like Camel-AI and other multi-agent LLM systems, and the team is committed to sharing insights and collaborating to accelerate progress. The group envisions producing significant outcomes, including co-authored publications, reusable tools, and metrics for wider applications. This initiative aims to set a foundation for transformative research in LLM multi-agent systems, with tools and findings applicable across various domains.