Multi-Agent Control: Probabilistic reasoning, optimal coordination, stability analysis and controller design for intelligent hybrid systems

The overall research objective of the network is to develop rigorous methods for analysis and design of Multi-Agent Control systems. Due to the interdisciplinary nature of this objective, the network has been structured to include expertise from relevant problem domains; probabilistic reasoning, optimisation, stability analysis, control theory and computing science. The specific design problems to be addressed are:

  1. To develop probabilistic reasoning methods for design that accommodate the inherent uncertainty in the system's knowledge of the state of the world. The work will build on new developments of computationally-intensive statistical inference tools, for modelling complex physical systems and human control behaviour.
  2. To develop tools for rigorously analysing the potentially very strong and safety critical interactions between the outcome of controller decisions and the dynamic behaviour of the overall system.
  3. To develop formal methods of design, which incorporate in a single framework, the design of the switching logic, co-ordination between multiple agents as well as optimisation of performance within given constraints on the overall system behaviour.

The emphasis in this network is to develop a theory to support the design of computer-controlled systems where performance and safety are crucial. The efficacy of the research results will be evaluated using a number of test-bed industrial applications (aerospace, automotive, process and renewable energy fields).

These applications will be supplied by a number of major European industrial companies, some of which are members of the network, and others that have expressed an interest in the scientific output of the network. Software tools developed during prototyping, as well as the scientific results, will be made available to the wider academic and industrial community.