实值多种群遗传算法求解动态规划问题

Optimization of Dynamic Systems Based on Real Value Multi-population Genetic Algorithms

研究利用遗传算法求解动态规划问题。实验采用实值多种群遗传算法，绕过复杂的数学推导，求解推车系统的最优控制序列u^＊（k）。在遗传算法迭代过程中，染色体采用实值编码、多种群、多目标并行搜索，并利用留优策略加速搜索收敛速度，求解得最优控制序列u^＊（k）。计算的目标函数值和数学解析解极值完全一致，证明了该方法的准确、高效。

The real value multi-population genetic algorithm is used to optimize the push-cart system. In this way, the complex mathematics inferential reasoning is bypassed, and the optimum control sequence u^＊ （k）of the push-cart system is obtained. In the genetic algorithm iterative process, multi-RVGA is used to compute the multi-goals parallel search, and search convergence rate is accelerated by using keeping the superior strategy. The goal function value based on the solution results in the optimum control sequence consist with the mathematics analytic selution, which proves the solution is accurate and highly effective.