A Novel Hybrid Algorithm Based on Lion Swarm Optimization and Differential Evolution Algorithm

Lion swarm optimization (LSO) algorithm, a novel and efficient meta-heuristic optimization technique, has received considerable attention and also has been applied to solve science and engineering optimization problems for outstanding performance. However, like some other typical swarm optimization algorithms, LSO may also trap individuals into local regions, resulting in premature convergence. Furthermore, how to balance contradictions between exploration and exploitation is also a critical issue that must be addressed in the LSO study. Hence, to address the aforementioned problems, a novel LSO variant algorithm called “hybrid lion swarm optimization (LSO) with differential evolution (DE)”, referred to as HLSOD, is proposed in this paper. First, LSO and DE are improved, respectively. For LSO, a nonhomogeneous cuckoo search strategy based on quantum mechanism is designed to enhance the exploitation capability. In contrast, a crossover strategy “crossover operator based on covariance matrix learning”, which not only can alleviate the algorithm’s dependence upon the coordinate system to some extent but also enhances the algorithm’s ability to tackle problems with high correlation between the variables, is employed to efficiently maintain the population diversity. For DE, the mutation strategy “DE/current-to-pbest” with optional external archive, which can use historical data to provide information of progress direction, is introduced as the mutation operator. The crossover strategy “crossover operator based on covariance matrix learning” is used for the crossover operator. Moreover, the adaptive parameter settings, which can automatically update the control parameters to appropriate values and avoid a user’s prior knowledge of the relationship between the parameter settings and the optimization problems’ characteristics, are applied for the parameter settings of the mutation and crossover operators. Thus, it is helpful not only for balancing contradictions between exploration and exploitation but also for improving the algorithm’s robustness. To enhance the algorithm’s convergence speed and maintain better candidate solutions, the greedy selection method is also adopted at each iteration of LSO and DE after the population is updated. Second, the improved LSO and the improved DE are hybridized to formulate HLSOD. A series of contrastive experiments are conducted on test functions from classic functions and CEC2017 test suite. Experimental results substantiate that the proposed HLSOD could achieve competitive or even better performance compared with other state-of-the-art algorithms. In addition, the application to four engineering design optimization problems further demonstrates the practicability and performance of HLSOD in tackling complex and diverse real-world optimization problem.