Towards More Effective Multi-objective Meta-Heuristics to Solve Complex Combinatorial Problems

本研究项目提出了多目标组合搜索领域中一些具有挑战性的思想的研究。该领域的大多数研究都是基于对单目标案例研究中获得的知识的回收，这激发了许多单目标技术向其多目标变体的扩展。其中许多是进化技术的扩展，如遗传算法、进化策略、粒子群优化等。将禁忌搜索和模拟退火等元启发式算法扩展到多目标变量。进化方法在多目标启发式搜索中受到了大多数关注，但我认为应该研究更广泛的元启发式技术。这里提出的研究主题代表了一个相当大的重点转变。具体来说，目标是构思更有效的多目标元启发式方法，以比当前技术水平更有效和高效的方式解决复杂的组合问题。本文旨在重新思考多目标元启发式的设计，在多目标范式中发展它们，而不是修改已知的单目标方法。

This research project proposes the investigation of a number of challenging ideas in the field of multi-objective combinatorial search. Most of the research in this area has been based on recycling knowledge acquired from research on the single-objective case and this has inspired the extension of many single-objective techniques to their multi-objective variants. Many of these are extensions from evolutionary techniques such as genetic algorithms, evolutionary strategies, particle swarm optimisation and others. There are some extensions of other meta-heuristics such as tabu search and simulated annealing to multi-objective variants. Evolutionary approaches have received most of the attention in multi-objective heuristic search but I believe that a wider range of meta-heuristic techniques should be investigated. The research themes proposed here represent a considerable shift in emphasis. Specifically, the aim is to conceive more effective multi-objective meta-heuristics to tackle complex combinatorial problems in a more effective and efficient manner than the current state of the art is capable of. This proposal aims to re-think the design of multi-objective meta-heuristics by developing them within the multi-objective paradigm instead of modifying known single-objective approaches.