基于核方法的移动机器人远程控制时延预测与分析*

构建了基于Internet的移动机器人控制平台，针对该平台中网络时延的非线性特性，提出了利用线性核和RBF核方法对网络时延序列样本进行回归建模和预测。比较了核方法、BP神经网络和RBF神经网络算法预测误差，表明在移动机器人远程控制中，核方法对网络时延序列预测具有更好的函数逼近能力和较高的预测精度，从而可根据预测时延调节移动机器人控制指令和状态信息的网络传输，保证系统的可靠性，提高移动机器人的控制性能。

Time delay prediction and analysis based on kernel method for mobile robot remote control

This paper built a mobile robot control platform based on Internet; with the consideration of the time-delay non-li-near characteristics of the platform proper, advanced the proposal of regression modeling and prediction to the time-delay series samples by means of linear kernel and RBF kernel methods. An error comparison was conducted between two kinds of predictions: the kernel method prediction and the BP neural network and RBF neural network method prediction. The result shows that, in the process of mobile robot control, compared with the neural network method, the adoption of the kernel method in the prediction of the network time-delay has better function approximation and higher prediction accuracy, so that, according to the predicted time-delay, it is possible to adjust the control instructions of the mobile robot and the network transmission of the status information to guarantee the system reliability and improvement of control performance of the mobile robot.