基于模型辨识的神经网络PID控制在经纬仪中的应用

Neural network PID control based on model identifier for theodolite

为了获得高精度和高速的位置控制,不确定参数如摩擦、惯量以及时延等都必须进行严格补偿,才能满足实时位置控制的需求.运用神经网络对非线性系统强大的自学习能力、记忆能力、计算能力以及各种智能处理能力,在控制系统中,能够学习和适应不确定性系统的动态特性,具有很强的容错性和鲁棒性,对经纬仪转台伺服系统进行在线辨识,得到系统的线性动态模型.利用此方法可以有效克服经纬仪的不确定因素,如电机参教的变化,负载转矩变化,摩擦的非线性变化等.在此基础上进行神经 PID 控制,PID 参数依据系统特性进行在线调整,从而达到更好的控制效果和更强的鲁棒性,得到了仿真和试验验证.

In order to obtain high precision and high speed response in the motion controling system, non-determinate variations,such as friction,torque,delay,ere,must be strictly compensated.Using neural network as a controller for the theodolite,which has the capabilities of self-learning,memory,extensive computation,and intelligent processing for complex,highly non-linear or indeterminate;the controller is often required to be sufficient robust in the presence of non-determinate variations in the parameters of the process.The neural network identifier on line can obtain the linear dynamic model of theodolite platform servo system,estimate the parameters of a controlled process,and modify a PID controller on line in order to meet the desired system performance and superior robustness At last,the results of simulation and experiment are verified.