十字路口类人行为的自动驾驶决策框架

Automatic driving decision algorithm for humanoid behavior at intersections

本文提出一个自动驾驶的决策框架,使得自主车辆能够安全、高效地通过十字路口。该框架首先使用数字地图来预测被观测自动驾驶车辆和行人的未来路径,使用预测的未来路径来识别潜在的威胁(行人)和碰撞区域;其次,在一个独立的分布式推理结构下,通过威胁测度、贝叶斯网络,系统可靠、稳健地评估潜在的威胁,利用这些信息,通过模拟器学习人类驾驶行为,为自动驾驶车辆确定适当的机动,以安全有效地行驶在十字路口。该算法具有很好的实验结果,能够成功地决定在行人通过道路之前或之后行驶。

This paper proposed an autonomous driving decision-making framework to enable autonomous vehicles to safely and efficiently pass through intersections.The framework used digital maps to predict the future paths of observed autonomous vehicles and pedestrians,and then used the predicted future paths to identify potential threats(pedestrians) and collision areas.Secondly,under an independent distributed reasoning structure,potential threats were systematically,reliably and stably assessed through threat measure and Bayesian network.Then the information is used to learn human driving behavior through the simulator to enable the appropriate movement for the autonomous vehicle to drive safely and effectively at the intersection.The algorithm shows good results and can successfully decide whether to drive before or after pedestrians pass the road.