Welcome
|
I am a Reader at the Department of Computer Science, School of Engineering and Applied Science, at Aston University, Birmingham, UK. I am a member of the ALICE research group of the Systems Analytics Research Institute, SARI. I graduated from the Department of Computing at Imperial College London and did my PhD at the Department of Electrical Engineering, Imperial College London, at the Intelligent Systems and Networks group, supervised by Prof Philippe De Wilde. |
Work in Brief
In most complex systems there is a mechanism that maps simple local rules to the macroscopic behaviour of these systems. We use agent-based modelling and simulation to investigate the following questions that arise from this fact:
- - Given the set of local interactions in a complex system (e.g. a society, an economy, an ecology) can we predict the large-scale patterns that emerge?
- - Given the large-scale patterns can we find the local interactions that lead to them?
- - Can we define local interactions that lead to the optimal large-scale patterns?
Recent research highlights
My work is in the area of artificial intelligence. I am particularly interested in applying machine learning techniques to intelligent agents and multi-agent systems, with a focus on societies and smart cities applications. Some recent research highlights:
- - Applying deep learning architectures for the first time for transfer learning in reinforcement learning contexts. Results (2019) show exceptional performance when used in newly-encountered settings.
- - Smart traffic: Real-time smart traffic management using deep reinforcement learning (2018) achieving large increase in traffic throughput and delay minimisation, along with ability to prioritise emergency vehicles.
- - Using Markov and Semi-Markov Decision Processes together with transfer learning for information fusion and decision-making. This technology is used within broker agents for energy trading and procurement, achieving outstanding results in the PowerTAC (2012-2015) and Supply Chain Management (2015) competition environments.
- - Coupling multi-agent systems and probabilistic approaches for mechanism design; devised a decentralised supply chain emergence mechanism utilising loopy belief propagation [1, 2, 3, 4], which outperforms existing approaches in the area.
- - Proposed the first trust and reputation system to fully rely on probabilistic modelling. Interactions of agents in an electronic marketplace (e.g. eBay) are modelled as a Hidden Markov Model. The approach significantly outperforms the state of the art.
- - Agent-based modelling for policy modelling, including models government policy for consumer attitudes (with BIS), effects of social influence and mobility to social violence.