An improved wild horse optimization algorithm based on reinforcement learning for numerical and engineering optimizations

Wild horse optimizer (WHO), inspired by the social life behavior of wild horses, is highly competitive in solving complex optimization problems. However, the original WHO is slow to converge in the later iterations, has low search accuracy, and is prone to fall into local optimum. In order to solve these problems, an improved wild horse optimization algorithm based on reinforcement learning and hybrid multi-strategy (IWHO) is proposed in this paper. Firstly, the initialization method of opposition-based learning is used to increase the population diversity and improve the quality of the initialized population. Secondly, the Q-Learning mechanism in reinforcement learning is introduced to establish a switching mechanism between grazing and mating behaviors of individual foals to guide the behavioral choices. Thirdly, a defense strategy is utilized to improve the algorithm’s optimization accuracy. Finally, last place elimination mechanism is adopted to eliminate the worst individual in each group and replaced it by random initialization to avoid the algorithm falling into a local optimum. The IWHO algorithm is tested on the CEC 2022 benchmark test suite and three practical engineering problems. The results show that IWHO has better performance than other algorithms.