有限感知信息下网联自动驾驶车辆编队的分布式控制方法

Control of connected autonomous vehicle formations, with which the safety, together with economy, comfort and efficiency of vehicles and the traffic system may get significantly improved, is one of the most important research area related to the connected and automated vehicles. And some research results have been applied in demonstrations. However, there are some limitations of current technologies, such as not good at adapting to the environment constraints, traffic scenes with the combination of connected and non-connected vehicles are not well-studied, and the expectation of perfect perception from the sensors in vehicular network. The main reasons contain the following items: the formation planning lacks of flexibility, the payoffs of the behavior control and decisions of vehicles are not analyzed, and the control algorithms are not optimized for systems with limited perception. To solve those issues, our research develops the thesis and methods of distributed control of connected autonomous vehicle formations with limited perception, which include: the method of the mission decisions of vehicles based on the environment constrains and vehicles’ payoff, adjusting the formation shape while assigning targets; Establish the payoff models of vehicles with discrete traffic scenes, study the cooperation of connected vehicles and competitions of non-connected vehicles; Analyze the effects of non-ideal communication channel, network topology variations and low precision feed-back, and optimize the control method. This research will provide both new thesis and methods in the area of distributed control of connected autonomous vehicle formations with limited perception.

网联自动驾驶车辆编队的控制方法是智能网联汽车领域的研究重点，可在提升车辆的安全性的同时，改善经济性、舒适性与行车效率，一些成果已得到初步示范应用。然而，现有技术存在对环境约束适应性差、未针对网联与网外车辆混合行驶的场景优化、对网联感知精度的依赖过高等不足。其主要原因是编队规划缺乏灵活性、对车间博弈收益考虑不足、未针对感知信息有限性优化控制算法。为此，本项目针对有限感知信息下网联车辆编队的分布式控制所涉及的理论方法及关键技术进行研究，内容包括：提出基于环境约束和车辆动力学的多车分布式任务决策方法，同步动态调节编队几何构型与车辆指派目标；建立离散交通场景车辆收益模型，提出基于网联车辆合作及网间车辆博弈的车辆行为控制决策方法；分析在非理想通信通道、变网络拓扑结构与低精度反馈下的优化控制方法。该研究可为国际上尚未解决的有限感知信息下的网联自动驾驶车辆编队的分布式控制问题提供新的理论方法与技术支撑。

网联自动驾驶车辆的编队协同决策与控制能有效提升交通效率与安全性，但现有算法存在场景普适性不高、无法处理混行博弈冲突场景、未考虑实际交通系统信息受限等问题。.本项目提出了面向多车交错编队的多车位置构型与指派方法，该方法覆盖包括一维道路及二维交叉路口常见交通场景，同时提升了交通安全性与道路利用率，且基于虚拟队列思想的子队列排序方法可保障车辆在复杂场景下的无冲突行驶；.所提出的基于车间博弈的意图识别模型，基于驾驶人理性行为假设，在混合博弈策略下估计理性驾驶行为，继而预测车辆的行驶轨迹，并建立动态交通势场，为自动驾驶提供决策与控制依据；.分析了数字通信通道质量与控制性能的映射关系，进而提出了丢包、时延、量化、截断等非理想通信反馈下网联控制系统的控制器设计方法；面向隐蔽型有界对抗信息攻击，提出了车辆纵横向控制器设计方法，保障了车辆的安全控制边界，提升了系统的鲁棒性和可靠性。.本项目构建了多车协同缩微平台、自动驾驶车辆平台与智能网联路侧系统，在其基础上开展了仿真与实车验证工作，初步测试结果证明了本项目提出方法的可用性。.已授权国家发明专利11项，申请14项。获得软件著作权2项。发表论文15篇，其中SCI收录2篇，EI收录12篇。培养博士生1人，在读博士5人，在读硕士1人。