基于反向学习的自适应α约束病毒种群搜索算法

中文摘要: 为了提高该算法求解约束优化问题的能力，提出一种新的约束病毒种群搜索算法。首先，提出自适应 α-level比较策略，以在算法的不同阶段充分利用可行个体与不可行个体的有效信息；其次，为了进一步提高算法求解约束优化问题的收敛速度和搜索精度，针对算法的病毒扩散行为，提出了结合反向学习机制的搜索方程，以提高种群多样性并加速全局收敛。对CEC2006中13个约束优化函数的对比仿真结果表明，本文算法在搜索精度、收敛速度以及稳定性方面，相比于 αSimplex算法、粒子群遗传算法算法、交叉人工蜂群算法算法以及约束改进差分进化算法算法具有明显优势。同时将该算法应用于人机协同实时航迹规划约束优化问题中，通过仿真实验并与利用约束改进差分进化算法对这一问题进行求解的方法进行对比，验证了本文算法在规划效率、规避威胁等方面的优越性。
Abstract:In order to improve the performance for solving constrained optimization problems, a novel constrained virus colony search (VCS) algorithm was proposed. Firstly, self-adaptive α-level comparison strategy was proposed to make full use of the feasible and infeasible solution at the different stages of the algorithm. Then, in order to improve the convergence rate and search ing precision for solving constrained optimization problems, the search equations based on opposition-based learning mechanism were designed for the process of virus diffusion in VCS. It was mainly used to improve the population diversity and convergence of the algorithm. Finally, the comparative experiments on thirteen CEC2006 benchmark functions showed that the proposed algorithm possessed more distinct advantages on searching accuracy, convergence rate and stability compared with αSimplex algorithm, particle swarm-assisted genetic algorithm (PSGA), crossover-based artificial bee colony (CBABC) and constrained optimization based on modified differential evolution (COMDE) algorithm. Meanwhile, compared with COMDE algorithm, the successful application in unmanned aerial vehicles (UAVs) cooperative path planning constrained problem verified the superiority of the proposed algorithm in the aspects of planning efficiency and threats avoiding.