Solving the VRP with probabilistic algorithms supported by constraint programming

Abstract

Among the human-being problems, there are those which carry with the exploration of a number of feasible and unfeasible possible solutions. These have been a challenge faced constructing specific strategies. Thus, techniques rising from several areas have been used. They can be complete or incomplete, with mathematical background or a more computational one, single techniques, greedy, constructive, etc. Most of these have proven to work fine when they are tailor-made solutions. Therefore, usually, the solution for a problem is totally useless for a different one or even for the same problem with new/different constraints. With this position paper, we claim a hybrid methodology combining probabilistic algorithms with a complete technique might be useful in order to separate the optimisation engine and the validation tool. This eases the possibility to have a general optimisation tool, with an acceptable minimal fine tuning, which includes neither the problem model nor validation jobs.

Peer Reviewed

Postprint (published version)