Enhancing Cuckoo Search Algorithm by using Reinforcement Learning for Constrained Engineering Optimization Problems

Cuckoo Search Algorithm (CSA) has been successfully applied to a range of various fields such as engineering, medical, and image processing. However, it typically suffers from a lack of effective exploration, loose diversity, and premature convergence. This paper aspires to develop a new version of CSA that is based on the features of Reinforcement Learning (RL) to enhance the research technique of CSA, which will be called CSARL. The performance of CSARL is evaluated by applying set of unimodal and multimodal benchmark functions. The results demonstrate that the CSARL outperforms the basic CSA, genetic, harmony search and krill heard algorithms, in terms of convergence speed, the diversity, and exploration search.