考虑车辆与人机混行环境耦合的L3级自动驾驶控制策略高效求解方法

针对复杂的人机混行环境，L3级自动驾驶车辆(L3-ADV)与有人驾驶车辆、非机动车、行人等混合行驶导致环境具有高机动性、强不确定性的特点，在考虑该不确定性的前提下如何保证L3-ADV安全、舒适、经济行驶已成为智能汽车行业共性科学难题。由于车辆与环境深度耦合机理引起的车车/车人博弈难题，及多目标协同优化难题，从而难以建立高效的行驶控制策略。因此，本项目提出复杂环境下的L3-ADV系统动力学实时控制方法，建立“车-环境”深度耦合机理、冲突行为博弈决策机制以及多目标动态博弈方法，并研制出L3-ADV域控制器样机。项目主要内容包括：1)设计基于信息物理融合的车辆行驶环境中时距感知方法；2)建立人机混行环境下的车辆潜在冲突消解与合作博弈决策模型；3)研究车辆运动规划及其实时高效解算方法；4)建立基于多目标实时博弈的L3级自动驾驶控制策略。研究工作将为L3级自动驾驶系统的实车应用提供理论支撑。

For the complex human-autonomous-driving mixed traffic environments, the L3-level autonomous driving vehicles (L3-ADVs) are driving together with the human-driving vehicles, non-motorized vehicles and pedestrians, which introduce the characteristics of high mobility and uncertainty into the environments. Considering this kind of uncertainties, how to ensure the safety, comfort and economic autonomous driving of L3-ADV has become a common scientific problem in autonomous automobile industry. Because of the game problem between the vehicle and other vehicles or pedestrians caused by the complex deep-coupled mechanism between vehicle and environment, the multi-objective coordination optimization problem, it is difficult to establish an efficient autonomous driving control strategy. Therefore, a real-time control method based on the system dynamics is proposed for L3-ADV in complex traffic environments in this project. The deep-coupled mechanism of vehicle-environment, game-theory-based decision-making mechanism of conflict behavior and multi-objective dynamic game methods are also established. Moreover, the domain controller is designed for L3-ADV. The main contents of the project include: 1) the medium-time-distance sensing method based on cyber physical system is designed; 2) the decision-making and cooperative game models of potential conflict resolution are established for the autonomous vehicle in human-autonomous-driving mixed traffic environments; 3) the vehicle motion planning and its real-time efficient calculation method are studied; 4) the L3-level autonomous driving control strategy is established based on multi-objective real-time game method. This research work will provide theoretical support for the real vehicle application of L3-level autonomous driving system.