A Hybrid Evolutionary Approach for Combinatorial Problems in Dynamic Environments

There is a growing interest in the use of evolutionary algorithms in time-varying environments where the information is revealed progressively with time to the decision maker. However, most existing research basically targets continuous optimization, while little work is directed to discrete optimization even though many real-world problems are both discrete and time-varying. This paper seeks to enhance the ability of genetic algorithms to track the optima shifting due to environmental changes: first, parameters of the genetic operators react to changes in the environment and to changes in the population diversity in order to persevere after obsolete convergence and overcome premature convergence. Second, multi-populations are introduced to cooperatively maintain the search diversity. Third, the algorithm is hybridized with local search heuristics for better tuning. The final algorithm is tested on dynamic combinatorial benchmark problems from the literature. Collectively, results from the conducted experiments favor the strategies proposed in this paper