基于交互式多模型算法的无人车车辆跟踪预测技术

Unmanned vehicle tracking prediction technology based on interacting multiple model method

针对无人车车辆路口通行中动态目标的跟踪预测,分析比较基于SICK与Velodyne两种激光雷达数据的车辆提取方法,参照Velodyne的提取方法,提出合适的激光雷达布局对路口环境中的动态障碍物(主要是车辆、行人)信息进行了提取.选取交互式多模型(IMM)算法对动态目标运动趋势进行预测,并对IMM算法进行优化,提出将局部路径规划的三次曲率多项式算法抽象为路径规划模型,作为IMM算法的滤波模型以替代常规的车辆运动模型作为滤波模型.验证实验结果表明基于路径规划模型的IMM算法在无人车车辆运动趋势的预测上具有更好的超前性与更高的预测精度.

To solve the tracking and predictability of unmanned car vehicle in dynamic objectives,the methods of vehicle detection by SICK and Velodyne are introduced and compared;the suitable laser radar layout on the dynamic obstacles in the road junctions environment（mainly is the vehicles,pedestrians） information extraction is carried out based on methods of vehicle detection by Velodyne.The interactive multi-model（IMM） algorithm is selected to predict the trend of the dynamic target motion.The IMM algorithm is optimized,the three times curvature polynomial algorithm of local path planning is Abstracted as path planning model;the conventional vehicle motion model is replaced by the IMM algorithm filter model as the filter model.The results show that the IMM algorithm using the path planning model has better advanced and higher precision in the predicting of behaviors of the vehicle.