摘要
针对在动态环境中自主车辆对于动态障碍物信息很难具有先验知识的问题,研究了动态贝叶斯网络模型对机动目标运动状态进行贝叶斯预测的推理机理,提出了一种基于贝叶斯预测进行自主车辆避障路径规划控制方法;该规划方法在VORONOI图法基础上,融合了对自主车辆和周围环境之间的位置关系的贝叶斯预测,一旦预定任务的动态环境发生重大变化,它可以产生机动目标沿某方向前进信息的预测先验知识,通过局部多次重规划生成避障路径,直至自主车辆完成既定任务;仿真实验证明了该规划控制方法可有效帮助自主车辆在不确定环境中实施避障策略。
In the dynamic surroundings, it＇ s a challenge to the autonomous vehicles （AVs） to acquire the prior knowledge about the surroundings. A Bayesian forecasting--based inference for mobile objects motion state using Dynamic Bayesian Network （DBN） model is studied in this paper. A novel Bayesian forecasting--based collision avoidance path planning for the AV using DBN is presented. This ap- proach fuses the forecasting--based relationship between the AV and the objects in the surrounding into VORONOI graph planning. Once the surrounding changes dramatically, this method may develop the prior knowledge complying with the object＇ s motion and generate the way- points for AV by local re--planning. Simulation results demonstrate this planning approach can valid the collision avoidance strategy for the AV in the uncertain surroundings.
出处
《计算机测量与控制》
CSCD
北大核心
2010年第7期1660-1663,共4页
Computer Measurement &Control
基金
国家自然科学基金重大研究项目(90205019)
关键词
自主车辆
不确定环境
避障
贝叶斯预测
autonomous vehicle
uncertain surrounding
collision avoidance
Bayesian forecasting
