52k (26k HWU, 26k BAE Systems): 01/05/2009 to 31/12/2009
Funded by the SEAS DTC (Systems Engineering and Autonomous Systems: Defence Technology Centre)
Two of the most recent broad developments for dealing with difficult real-world optimisation problems are the concepts of multiobjective optimisation and probability collectives. Multiobjective optimisation has been around for a while, but it is relatively recently that algorithms have emerged that enable problems to be handled efficiently and effectively in their true multiobjective form, rather than attempting to solve a simplifed (some say `brutalised') single-objective version of the problem. Meanwhile, many different cognate communities are converging upon optimisation algorthms that are centred on the use of probablistic models of good solutions, and iteratively sample from and update the model: one of the more recent and sophisticated such methods is probability collectives, which has been shown to have superior performance on several problems. In this project we start to investigate combining these developments, and will develop multio-objective probability collectives, and ultimately test them on one or more real-world problems such as autonomous vehicle path planning.
Project Team:
Prof. David Corne (Heriot Watt), Dr Andrew Waldock (BAE Systems)