A Hybrid Genetic Algorithm for the 0-1 Multiple Knapsack Problem

A hybrid genetic algorithm based in local search is described. Local optimisation is not explicitly performed but it is embedded in the exploration of a search metaspace. This algorithm is applied to a NP-hard problem. When it is compared with other GA-based approaches and an exact technique (a branch and bound algorithm), this algorithm exhibits a better overall performance in both cases. Then, a coarse-grain parallel version is tested, yielding notably improved results.