基于PSO-NARX网络的司机驾驶行为分析方法

Driver Driving Behavior Analysis Method Based on PSO⁃NARX Network

舒适性、准时性、节能性等是衡量高速铁路自动驾驶水平的重要指标,通过不断学习优秀司机的驾驶行为,可以优化列车自动驾驶性能,促进高速铁路自动驾驶技术的发展。基于现场列车运行数据,提出一种带有外部输入的非线性自回归(NARX)网络的列车司机驾驶行为分析方法。该方法构建了具有时序特征的NARX网络模型,并选取多项影响司机决策的参数作为输入,利用粒子群优化算法(PSO)确定网络的权重和阈值,对下一时刻列车运行情况进行预测。仿真结果表明:本文提出的PSO-NARX网络分析模型的预测效果优于前馈型神经网络(BP)、PSO-BP、NARX,相比于BP算法,迭代步数降低了373步,误差降低了8382%,相关系数达到了90117%。通过此预测,可以优化列车的自动驾驶设备性能指标,保障列车准时的同时,提高了乘客乘坐的舒适性。

Comfortability,punctuality and energy efficiency are important indicators for evaluating the level of automatic driving of high-speed trains.Through continuous learning from the driving behaviors of excellent drivers,the automatic driving performance of trains can be optimized,which can promote the development of high-speed train automatic driving technology.Based on on-site train operation data,this paper proposed a nonlinear auto-regression with exogenous inputs(NARX)network analysis method for analyzing the driving behavior of train drivers.The method constructed an NARX network model with temporal characteristics,and selected multiple parameters that affect driver decisions as inputs.At the same time,the weight and threshold of the network were optimized using particle swarm optimization(PSO),to pre-dict the operation situation of the next train.The results show that the PSO-NARX network analysis model proposed in this paper performs better than the back propagation(BP)neural network,PSO-BP,and NARX.Compared with BP al-gorithm,the iteration steps are reduced by 373 steps,with an error reduction of 8382%and a correlation coefficient of 90117%.Through this prediction,the performance indicators of the automatic driving equipment of the train can be op-timized to ensure the punctuality of the train,while improving the comfort of passengers.