The Stochastic Agents Group at Imperial College London is involved in research on modeling and application of agent technologies for solution of a wide range of problems of current and future interest. We have a particular interest in novel agent models based on a stochastic framework for efficient simulation and analysis of large-scale agent systems. We are actively involved in the following research projects with Government and Industry:

SATURN, Self-organizing Adaptive Technology underlying Resilient Networks

We develop systems that aim at improving the resilience of critical infrastructures. To this effect, we investigate distributed algorithms inspired by opportunistic communications and epidemic routing. We evaluate our proposed methods inside large scale simulations of urban areas, using the DBES. Our work also includes theoretical analysis and modeling of communication networks in order to prevent network attacks.

Former projects

ALADDIN, Autonomous Learning Agents for Decentralised Data and Information Networks
In collaboration with British Aerospace Systems and EPSRC

We developed methods to compute the equilibrium behaviour of large agent teams which interact in a stochastic environment. We studied the effect of heuristic algorithms and of locally optimal solutions on large scale system behaviour. To this effect, we developed models based on Chapman-Kolmogorov equations and on optimal control. System oscillations and their stabilisation are considered. The research was applied to issues related to emergency management and control of its effects.
We have also developed a Distributed Building Evacuation Simulator (DBES), which addresses the simulation of emergency response scenarios.

Future Data Fusion Systems (2003-2006) and C2 Agents (2006-2009)

This project developed methods that helped to automate decision making processes for adversarial agents by fusing different types of information, such as objectives, environmental characteristics, capabilities of different agents and teams, and threats. These methods can also be used to assess the efficacy of different decisions based on the measurable technical characteristics and capabilities of the agent teams. The impact of the Network Enabled Capabilities (NEC) of the agent teams was also included in the decision process.