摘要
为了提供复杂道路上汽车自动驾驶的目标速度,提出了基于前视轨迹曲率的速度决策算法:首先决策出前视断面的轨迹点,算出每个点位的轨迹曲率并将其作为输入数据;然后从时间最省、驾驶最舒适、定速巡航以及混合模式中选择其一作为决策目标;在速度界限以及纵向/侧向舒适性约束下进行滚动时域优化,决策出前视断面上的期望速度值,随着车辆的行驶前视断面依次向前滚动,最终得到沿行驶距离变化的速度曲线.用复杂山区道路的实测数据验证了模型的精度,以2条公路和1条赛道作为仿真算例,结果表明:(1)通过组合不同的轨迹决策目标和速度决策目标,能够得到多种驾驶模式的速度曲线,进而模拟出多种实际驾驶行为;(2)由于前视轨迹是在通道边界内生成,弯道半径、转角、回旋线、路宽、偏转方向等用于确定通道边界的道路变量,都会改变轨迹特性进而影响行驶速度,因此,本文算法能够适应复杂的道路几何条件.
To provide the target speed profile for automatic driving of cars, a speed determination algorithm was presented based on the curvature of pre-view trajectory. Firstly, the position of trajectory on several preview cross-sections was determined and the curvature of each trajectory point was calculated and used as the input data of speed optimization. Secondly, selecting an objective function from minimum traveling time, driving comfort, minimum deviation from the cruise speed and a mixed mode, a speed decision was made under constraints of the lateral/longitudinal driving comfort and the maximum/minimum speed limit using a moving horizon algorithm. With the rolling of preview crosssections with the forward movement of ears, the desired speed profiles along the distance traveled can be obtained. Finally, a field test was carried out on a complex mountainous highway to validate the proposed model, and examples of trajectory and speed decision of two mountainous highways and an F1 track was illustrated. The results show that: (1) speed profiles of different driving patterns were obtained though the combination of different modes of trajectory control and speed control, which could simulate the various driver behaviors; (2) as the simulated trajectory is generated within the road geometry, factors used for determining the edge of the roadway, such as radius, deflection angle, spirals, pavement width, deflecting direction can all affect the trajectory and then affect the driving speed, therefore the model is applicable to the highways with complex geometrical shapes.
出处
《西南交通大学学报》
EI
CSCD
北大核心
2015年第4期577-589,共13页
Journal of Southwest Jiaotong University
基金
国家自然科学基金资助项目(51278514)
高等学校博士学科点专项科研基金资助项目(20135522110003)
关键词
人车路系统
智能驾驶
目标速度
运行速度
速度决策
驾驶行为
驾驶员模型
human-vehicle-road system
intelligent driving
target speed
operating speed
speed decision
driver behavior
driver model