| 基于贝叶斯概率估计的智能电动车动态目标避障算法 |
| |
| 引用本文： | 盛鹏程,曾小松,罗新闻,马金刚,戎辉,卞学良.基于贝叶斯概率估计的智能电动车动态目标避障算法[J].中国公路学报,2019,32(6):96-104. |
| |
| 作者姓名： | 盛鹏程 曾小松 罗新闻 马金刚 戎辉 卞学良 |
| |
| 作者单位： | 1. 河北工业大学 机械工程学院, 天津 300401;2. 邢台职业技术学院汽车工程系, 河北 邢台 054000;3. 工业和信息化部装备工业发展中心, 北京 100846;4. 中国汽车技术研究中心 汽车工程院, 天津 300300 |
| |
| 基金项目： | 国家博士后基金项目（2017M611180）；国家重点研发计划项目（2017YFB0102500）；河北省高等学校青年基金项目（QN2019094） |
| |
| 摘 要： | 为了实现智能电动车在中汽中心智能网联示范基地内的动态避障，首先将直角坐标系与曲线坐标系进行转换，构建以参考路径的弧长s为横坐标，横向偏移距离q为纵坐标的曲线坐标系；其次，在曲线坐标系中利用三次多项式生成满足初始位姿与子目标点位姿的候选路径，同时对标准化常量的似然函数进行定义，在此基础上利用贝叶斯定理对每条候选路径的危险等级进行概率估计；在动态避障过程中，借鉴速度障碍法对碰撞威胁进行实时检测，并建立最短避障时间和安全距离的数学模型来实现高效的动态避障，最后对行人占用车道行走与横穿马路2种典型场景进行动态避障试验。研究结果表明：在曲线坐标系中，通过横向偏移距离能够便捷地建立起一系列候选路径，克服在直角坐标系中寻找移动子目标点这个难题；在寻找安全路径方面，由于智能电动车工作环境的不确定性，利用贝叶斯定理对候选路径危险等级进行概率计算的方法可靠性更高，速度障碍法与避障数学模型的结合满足碰撞危险检测的实时性和动态避障的高效性要求。试验结果表明：采用曲线坐标系中的动态避障算法对行人占用车道和横穿马路2种场景进行了有效的避障，在路径选择上符合实际驾驶习惯，达到了智能网联示范基地动态避障的要求。
|
| 关 键 词： | 汽车工程 动态避障 速度障碍法 贝叶斯概率 曲线坐标系 |
| 收稿时间： | 2018-08-11 |
Multi-objective Dynamic Obstacle Avoidance Algorithm of Intelligent Electric Vehicles Based on Bayesian Theory |
| |
| SHENG Peng-cheng,ZENG Xiao-song,LUO Xin-wen,MA Jin-gang,RONG Hui,BIAN Xue-liang.Multi-objective Dynamic Obstacle Avoidance Algorithm of Intelligent Electric Vehicles Based on Bayesian Theory[J].China Journal of Highway and Transport,2019,32(6):96-104. |
| |
| Authors: | SHENG Peng-cheng ZENG Xiao-song LUO Xin-wen MA Jin-gang RONG Hui BIAN Xue-liang |
| |
| Affiliation: | 1. School of Mechanical Engineering, Hebei University of Technology, Tianjin 300401, China;2. Automotive Engineering Department, Xingtai Polytechnic College, Xingtai 054000, Hebei, China;3. Ministry of Industry and Information Technology Equipment Industry Development Center, Beijing 100846, China;4. Automotive Engineering Research Institute, China Automotive Technology and Research Center, Tianjin 300300, China |
| |
| Abstract: | In this study, a dynamic obstacle avoidance algorithm based on the Bayesian theory was proposed for an intelligent electric vehicle in the CATARC intelligent network demonstration base. Firstly, the Cartesian coordinate system and the curvilinear coordinate system were transformed, and an arc length with the reference path as the abscissa and the horizontal offset distance as the ordinate of the curvilinear coordinate system was thereby constructed. A cubic polynomial in the curvilinear coordinate system was then employed for generating candidate paths that satisfy the initial pose and the sub-target pose, and the likelihood function of the normalization constant was simultaneously defined. The Bayesian theory was thereby applied for estimating the probability of each candidate path. A mathematical model for the shortest obstacle avoidance time and corresponding safety distance was also established for realizing efficient dynamic obstacle avoidance. Furthermore, a dynamic obstacle avoidance experiment was conducted for two typical scenarios involving pedestrians occupying lanes and crossing roads. The obtained results indicate that a series of candidate paths can be established by incorporating the lateral offset distance in the curvilinear coordinate system, which in turn can overcome the difficulty of locating moving sub-targets in the Cartesian coordinate system. While searching safe paths, owing to the uncertainty of the operating environment of an intelligent electric vehicle, the Bayesian theory is found to be more reliable for probability calculation of candidate path hazards. Combining the velocity obstacle method with an obstacle avoidance mathematical model can help achieve real-time performance of collision risk detection and can meet the high efficiency requirements of dynamic obstacle avoidance. Hence, the proposed dynamic obstacle avoidance algorithm can be applied in the curvilinear coordinate system to effectively avoid obstacles for pedestrians occupying lanes and crossing roads. Furthermore, the proposed route selection process meets the requirements of actual driving habits, and the requirements of dynamic obstacle avoidance in the intelligent network demonstration base can thus be achieved. |
| |
| Keywords: | automotive engineering dynamic obstacle avoidance velocity obstacle method Bayesian probability curvilinear coordinate system |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《中国公路学报》浏览原始摘要信息 |
|
点击此处可从《中国公路学报》下载全文 |
|
