说明：  Parsopoulos等采用罚函数法，利用非固定多段映射函数对约束优化问题进行转化，再利用PSO算法求解转化后问题，仿真结果显示PSO算法相对遗传算法更具有优越性，但其罚函数的设计过于复杂，不利于求解；Hu等采用可行解保留政策处理约束，即一方面更新存储中所有粒子时仅保留可行解，另一方面在初始化阶段所有粒子均从可行解空间取值，然而初始可行解空间对于许多问题是很难确定的；Ray等提出了具有多层信息共享策略的粒子群原理来处理约束，根据约束矩阵采用多层Pareto排序机制来产生优良粒子，进而用一些优良的粒子来决定其余个体的搜索方向。
(Parsopoulos etc. penalty function method, the use of non-fixed multi-stage mapping function constrained optimization problem for transformation, reuse PSO algorithm for solving the problem after the conversion, the simulation results show that the relative genetic algorithm PSO algorithm has more advantages, but its design is too complicated penalty function is not conducive to solving Hu retention policies such as the use of feasible solutions processing constraints, on the one hand to retain only feasible solution when updating the store all particles, on the other hand all the particles are in the initialization phase values ​ ​ from the feasible solution space, however, the initial feasible solution space For many the problem is difficult to determine Ray and put forward the principle of a multi-layer particle swarm information sharing strategies to deal with constraints, according to the constraint matrix using multi Pareto superior sorting mechanism to generate partic)