基于改进人工势场法的路径规划决策一体化算法研究

Integration Algorithm of Path Planning and Decision-making Based on Improved Artificial Potential Field

摘要:

车辆路径规划与决策算法是无人驾驶汽车的重要研究方向之一。现有的路径规划与路径跟踪决策算法中规划层与决策层存在时滞现象，往往会引起检测信息与真实行驶环境信息的偏差，使得规划的局部路径不能反映当前真实状态，是无人驾驶汽车安全行驶的不稳定因素。本文综合考虑了环境交通参与者与车辆自身运动学特征，建立了横纵向安全模型，对车辆目前行驶环境的风险特征进行了综合评估。在行驶环境特征与车辆动力学特征的基础上对传统人工势场法进行了改进，设计了基于虚拟力的局部路径规划与控制决策一体化算法，提升了算法在复杂动态环境下控制的可靠性。最后，利用Carsim/Simulink建立了联合仿真环境，分别对传统路径规划算法、路径跟踪算法与本文提出的路径决策规划一体化算法在典型工况下进行仿真。仿真结果表明，该算法能减小路径规划决策环节的时滞影响，为复杂动态环境下的无人驾驶车辆提供更加合理的控制方法。

Abstract:

Path planning and decision-making algorithm is one of the most important research directions of driverless vehicles. However, the delays of the path planning and decision-making algorithms could lead to the inconsistency between sensor information and real driving environment, introducing negative effects on the ability to avoid dangerous state. Classified longitudinal model was established by considering the expected route, kinematics characteristics of environmental traffic participants and vehicles to determine the safety condition of vehicle. Also, a lateral safety space model was established to determine whether it was safe to change lane. Based on the safety model, combining the environmental and vehicle dynamic characteristics, an integrated algorithm of local path planning and decision-making algorithm was provided to improve the performance of the algorithm in complex dynamic environment. In the model, the influence of environmental information was represented with artificial force such as global planning gravitation, lane changing gravitation, forward obstacle repulsion and sensor occluded scenes repulsion. Gravitations represented attractive factors’ influence and repulsions represented repulsive factors’ influence of environment. Finally, co-simulations based on Carsim/Simulink was established to analyze the delay of traditional algorithm and algorithm proposed under various typical conditions. Results showed that the proposed algorithm can reduce the time-delay effect of path planning and decision-making, and provide better control for unmanned vehicle control in complex dynamic environment.