| 基于状态空间采样的高速公路智能网联车辆轨迹动态规划 |
| |
| 引用本文： | 郭烈,王旭,胥林立,秦增科.基于状态空间采样的高速公路智能网联车辆轨迹动态规划[J].中国公路学报,2022,35(12):253-267. |
| |
| 作者姓名： | 郭烈 王旭 胥林立 秦增科 |
| |
| 作者单位： | 1. 大连理工大学 汽车工程学院, 辽宁 大连 116024;2. 大连理工大学 宁波研究院, 浙江 宁波 315016 |
| |
| 基金项目： | 国家自然科学基金项目(51975089)；国家重点研发计划项目(2018YFE0197700)； 辽宁省自然科学基金项目(2021-MS-127)；宁波市重大科技任务攻关项目(2022Z017)* |
| |
| 摘 要： | 为实现智能网联车辆在高速公路动态行车环境下的轨迹实时规划，提出一种基于状态空间采样的轨迹动态规划方法。首先，以安全性为原则选取主车当前行驶的理想车道。基于Frenet坐标与笛卡尔坐标的转换关系，建立车辆运动横、纵向解耦的独立积分系统。将高速公路常见的行驶状态分为车道保持与定速巡航、变道以及前车跟随3类，预测主车行驶车道并针对3类行驶状态分别设计轨迹终端的目标配置方法。然后，利用多项式函数生成连接初始配置和目标配置的多条待选轨迹。构建考虑轨迹偏离理想车道程度、始末速度变化、规划周期和轨迹舒适性的综合损失函数，结合速度、加速度、曲率检查来评价各条待选轨迹的成本并进行排序。最后，预测车辆的横、纵向运动轨迹并构建一种胶囊形的车辆虚拟安全边界，通过碰撞检测，确定主车的最优轨迹，设置动态规划触发条件及时更新最优轨迹并避免过度规划浪费资源。研究结果表明：提出的算法能满足高速公路场景的动态规划需求；通过对轨迹规划周期、虚拟安全边界、动态规划时间间隔等关键参数的分析与优化，主车的横摆角速度范围稳定在-0.1~0.15 (°)·s-1，横向加速度范围稳定在-0.16~0.32 m·s-2，跟踪参考轨迹的最大误差不超过0.022 m，提出的算法能规划出具有高安全性、稳定性和舒适性的轨迹。
|
| 关 键 词： | 汽车工程 轨迹动态规划 状态空间采样 智能网联车辆 损失函数 碰撞检测 |
| 收稿时间： | 2021-12-10 |
Dynamic Trajectory Planning of Intelligent Connected Vehicle in Expressway Environment Based on State Space Sampling |
| |
| GUO Lie,WANG Xu,XU Lin-li,QIN Zeng-ke.Dynamic Trajectory Planning of Intelligent Connected Vehicle in Expressway Environment Based on State Space Sampling[J].China Journal of Highway and Transport,2022,35(12):253-267. |
| |
| Authors: | GUO Lie WANG Xu XU Lin-li QIN Zeng-ke |
| |
| Affiliation: | 1. School of Automotive Engineering, Dalian University of Technology, Dalian 116024, Liaoning, China;2. Ningbo Institute, Dalian University of Technology, Ningbo 315016, Zhejiang, China |
| |
| Abstract: | A dynamic trajectory planning method based on state space sampling was developed to realize real-time trajectory planning of intelligent connected vehicles in the dynamic driving environment of an expressway. First, the current ideal lane of the host vehicle was selected based on the safety principle. An independent integral system with horizontal and vertical decoupling of vehicle motion was established based on the conversion relationship between Frenet and Cartesian coordinates. The common driving states of the expressway were divided into three types: lane keeping and cruise control, lane changing, and car following. The future lane on which the host vehicle will drive was predicted, and the target configuration methods of trajectory terminals were designed for the three types of driving state. Then, a polynomial function was used to generate multiple candidate trajectories connecting the initial and target configurations. A comprehensive loss function was constructed that considers the degree to which the trajectory deviates from the ideal lane, as well as changes in speed from beginning to end, the planning period, and trajectory comfort. The cost of each trajectory was evaluated and sorted by combining speed, acceleration, and curvature checks. Finally, the horizontal and vertical trajectories of the vehicle were predicted, and a capsule-shaped virtual security boundary was constructed. The optimal trajectory of the host vehicle could be determined through collision detection. Results show that the algorithm can meet the dynamic trajectory planning requirements of expressways. Through the analysis and optimization of the critical parameters, such as the trajectory planning period, virtual security boundary, and dynamic planning interval, the yaw rate range of the host vehicle was stabilized at -0.1-0.15 (°)·s-1, the lateral acceleration range was stabilized at -0.16-0.32 m·s-2, and the maximum error of the reference trajectory tracking was not more than 0.022 m. The algorithm can be used to plan trajectories with high safety, stability, and comfort. |
| |
| Keywords: | automotive engineering dynamic trajectory planning state space sampling intelligent connected vehicle loss function collision detection |
|
| 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 |
|
