A reinforcement learning assisted evolutionary algorithm for constrained multi-task optimization

Yufei Yang, Changsheng Zhang, Bin Zhang, Jiaxu Ning

Abstract: Highlights•A reinforcement learning-based evolutionary algorithm designed for constrained multi-task optimization.•A multi-population method is employed to facilitate populations traversal through infeasible regions.•An adaptive operator selection strategy based on Q-Learning with upper confidence bound.•An dimension-based knowledge transfer is employed for constrained multi-task optimization.•Comprehensive experiments confirming the effectiveness and superiority of the proposed algorithm.