基于精英进化导向的多目标PID参数优化

在多目标优化问题中,决策者必须对Pareto前沿的众多非劣解做出选择.本文将决策偏好融入Pareto优化过程,提出一种基于精英导向机制的多目标遗传算法,根据决策偏好选择Pareto最优解为精英,利用无损有限精度法和归一增量距离保持种群多样性,通过多种群进化机制将决策偏好的影响传播到整个种群.该方法成功应用于自动导引车(AGV)伺服系统的PID参数优化,可根据决策偏好快速有效地定向搜索Pareto最优解,保证伺服控制达到路径跟踪要求的速度响应性能.

Multi-objective optimization for PID parameter based on elitist-evolution guidance

For multi-objective optimization problems, a decision-maker must choose one solution from many nondominated ones in Pareto front. Decision preferences are introduced into Pareto optimization in this paper, and a multiobjective genetic algorithm based on elitist-guidance mechanism is presented. Elitists are selected from Pareto optimal solutions according to decision-making preferences. The lossless-finite-precision method and the normalized incrementdistance are proposed to keep the population diversity. The effect of decision-making preferences is spread among the entire population by using the multi-population evolution mechanism. This approach is applied successfully to PID parameter optimization of automated-guided-vehicle(AGV) servo system, which can make a fast, effective and directional search for Pareto optimal solutions according to decision-making preferences, and ensures the servo control for achieving the velocity-response performance required by path tracking.