【摘要】 针对汽车操纵逆动力学的研究现状,采用实神经网络代数算法,对汽车逆动力学进行研究。根据二自由度汽车开环系统模型的运动微分方程和双移线道路模型,进行"正问题"的求解,将获得的样本输入实神经代数网络进行训练。将训练过程与BP网络算法的结果进行对比,说明代数算法的优越性。另取一组仿真值输入训练成功的代数网络进行识别,将得到的汽车方向盘转角与仿真结果进行对比,验证网络精度。最后进行实车试验,将试验数据与神经网络识别结果进行对比,再次验证其准确性。更多还原

【Abstract】 Based on the present research of the vehicle handling inverse dynamics, the algebraic algorithm of real neural network is used to study the inverse dynamics. According to the motion differential equations and double line road model of two degrees of freedom vehicle open-loop system model, to solve the "forward problem", the samples achieved are input into real neural network for training. The training process of the algorithm network is compared with BP, and the result shows the superiority of algebraic algorithm. Another set of simulation which can identify successful network is input, the achieved car steering wheel angle is compared with the simulation results, so as to verify the accuracy of network. At last, real vehicle experiment is carried out, the test data is compared with the neural network identification result, again to verify its accuracy.更多还原