优化PID与神经PID控制主动悬架的性能对比研究

Comparative study on performance of active suspensions based on optimized PID and neural PID control

为主动悬架选择一种更可行的控制方法,对PID与神经PID控制主动悬架进行了优化后的性能对比研究。基于1/4车二自由度主动悬架模型,利用遗传算法以悬架二次型性能指标为目标函数,分别对PID控制主动悬架的增益系数与神经PID控制主动悬架的初始权值和学习效率进行了优化。优化结果显示:优化后的PID控制主动悬架的综合性能较神经PID控制主动悬架略优。出现上述结果的原因在于:当神经PID控制主动悬架的学习效率等于零时则退化成PID控制主动悬架,学习效率不等于零则导致神经PID控制主动悬架的实时PID权值偏离了最优的PID权值。此外凸块路面输入下的仿真也显示优化PID的鲁棒性也略优于优化神经PID。因此,选择算法较复杂的神经PID对主动悬架进行控制是没有必要的。

To choose a feasible control strategy for active suspension,the performance comparison between optimized PID and neural PID control for active suspension is presented.Based on a quarter-vehicle active suspension model,the quadratic performance index for active suspension shall be regarded as target function by applying genetic algorithm,and the gain coefficients of PID control and the initial weights and learning efficiency coefficients of neural PID shall be optimized respectively.The optimization result shows the comprehensive performance of optimized PID control active suspension is a little better than that of optimized neural PID control one.The reason lies in the following：the optimized neural PID control will degenerate into an optimized PID one when the efficiency coefficients of optimized neural PID control are set as zeros;when the learning efficiency coefficients are not equal to zeros,real-time PID weights of neural PID control will deviated from the optimized PID weights and lead to the above result.Simulation in bump road input demonstrates robustness of active suspension based on optimized PID is better than the one of active suspension based on optimized neural PID too.Therefore,it is not necessary for active suspension to choose more complex neural PID control.