Multi-agent based evolutionary algorithm for dynamic knapsack problem

Dynamic optimization problems have attracted widely interest lately as many real world problems are indeed dynamic. The dynamic knapsack problems area class of well known dynamic testing problems, which can bridge the gap between the very complex, hard to understand real-world problems and the too simple toy problems because many real problems can be described by the dynamic knapsack problems. Natural organisms have always exhibited good adaptability to changing environments, which bring evolutionary algorithm a great source of inspiration to address dynamic optimization problems. In this paper, a multi-agent based evolutionary algorithm (MAEA) is introduced to solve dynamic knapsack problems. The agents simulate living organism features and co-evolve to find optimum. And random immigrants scheme is used to maintain the diversity of the population. Simulation experiments on a set of dynamic knapsack problems show that the proposed MAEA can obtain a better performance in comparison with several peer genetic algorithms.