车辆单神经元模型参考自适应控制算法研究

Study on Single Neuron Model Reference Adaptive Control Algorithm for Vehicles

针对智能车辆油门控制系统,提出了一种单神经元模型参考自适应控制算法.首先通过实验研究获得油门控制系统的传递函数,再以该函数获得的数学模型为依据设计了自回归滑动平均模型(NARMAX)神经网络,并对系统输出进行离线辨识和在线预测.采用免疫模糊思想改进二次型单神经元控制算法,构建基于NARMAX神经网络预测的模型参考自适应控制系统,定义了一种评价车辆纵向运动的目标函数,采用浮点遗传算法寻找各控制器的最优值.仿真结果表明,NAR-MAX神经网络可辨识和预测车辆油门系统的动力特性,与免疫模糊和二次型单神经元算法相比,单神经元模型参考自适应算法的阶跃响应速度显著提高.

Focusing on the throttle control system of intelligent vehicle, a model reference adaptive SN-PID controller is presented. Firstly, we obtain the transfer function of a throttle control system by experimental method and design a NARMAX neural network based on the above model, and then the off-line identification and online prediction are carried out. Secondly, the linear quadratic SN-PID algorithm is improved by using immune fuzzy ideas. A model reference adaptive system based on the prediction model of NARMAX neural network is created and an objective function is defined to evaluate the vehicle longitudinal motion, and then the optimal value of all the adjustable parameters in the above controllers is found by the float genetic algorithm. Finally, a digital simulation is carried out to compare the dynamic performance of improved SN-PID controller with the classical one. Results show that the proposed NARMAX neural network is capable of identifying and predicting the output of throttle control system, and the step response velocity of the proposed SN-PID controllers is remarkably faster than that of immune fuzzy and classical ones.