无级转向履带车辆驾驶员转向操控行为建模与预测
Modeling and Prediction of Driver's Steering Control Behavior for Stepless Steering Tracked Vehicle
杜云生 1王文硕 1魏源 1陈慧岩1
作者信息
- 1. 北京理工大学 机械与车辆学院,北京 100081
- 折叠
摘要
基于现役有人驾驶履带车辆进行无人化改造是无人履带车辆研制的有效技术途径之一.为实现某现役履带车辆无人化线控技术改造,设计车辆数据采集系统,以实现车辆状态数据与驾驶员操控数据的同步采集.依据人类驾驶员的操控数据,基于高斯混合模型对驾驶员转向操控行为进行聚类分析,并建立转向操控行为概率模型.基于不同转向类别,以车辆行驶速度和航向角偏差为模型输入,以静液马达摆臂转角为转向操控量预测真值,对驾驶员转向操控预测模型进行训练,实现对驾驶员转向操控行为准确预测,并在真实越野场景下进行验证.试验结果表明,所提转向操控预测模型可以准确地预测驾驶员转向操控.
Abstract
Reforming the current manned tracked vehicles into the unmanned tracked vehicles is one effective way to develop the unmanned tracked vehicle.To this end,a data acquisition system is designed to collect the vehicle state and driver's operation data.Based on the collected data,the driver's steering behaviors during the tracked vehicle's operation are clustered and analyzed based on the Gaussian mixture model(GMM),and a steering control behavior model is established.Based on different steering categories from GMM,a prediction model for the driver's steering operation is trained by taking the running speed and steering angle deviation of vehicle as model inputs and the turning angle of hydraulic motor swinging arm as the predictive truth value of steering control.The proposed predsiction model is used for the statistical modeling and prediction of driver steering behavior in the in real cross-country environment.Experimental results show that the proposed steering control model can predict the driver's steering behavior accurately.
关键词
履带车辆/操控行为/行为预测/高斯混合模型/随机森林Key words
tracked vehicle/steering behavior/behavior prediction/Gaussian mixture model/random forest引用本文复制引用
出版年
2024
NETL
NSTL
万方数据