The autonomous systems research group is part of the EITAMS project. Please refer to the project webpage for further information.

A large number of algorithms for solving optimisation problem has been developed. However, solving any optimisation problem requires selecting a suitable algorithm with a suitable configuration. Different methodologies have been followed to automate deciding on the right algorithm and the right configuration. Hyper-heuristics approaches may not compete with state-of-the-art problem-specific approaches because they focus on the performance of the optimisation algorithms rather than the problem. On the other hand, most of Rice’s features-based model approaches depict the Algorithm Selection (AS) and the Algorithm Configuration (AC) problems as two separated problems. This separation can affect the algorithm efficiency or narrow the range of problems, on which an approach can be applied.

This research project aims to investigate extending Rice’s framework for solving the AS problem to enable producing search results that comparable to that produced by state-of-the-art problem-specific approaches. By incorporating the algorithms’ features into the features-based model, powerful and more general, solver can be developed without losing domain-independence.