智能网联汽车轨迹规划与通信协作的联合优化理论与关键技术研究

Taking the requirement of safety driving and the limit of network resource as constraints, the controlled variables of trajectory planning and communication as the objects, maximizing performance of trajectory planning and communication as the objectives, this research aims at the best solution towards cooperative trajectory planning through progressively establishing the dependence-relation graph, the single-vehicle scene perception and multiple-vehicles-based game dynamics model and solving them. 1) Using Graph Description Language (DOT), we describe the inter-dependent and inter-constrained relations among context-awareness, trajectory planning, communication and decisions. 2) We employ Partially Observed Markov Decision Process (POMDP) to model inter-relations between single vehicle and sensing context against error and inaccuracy of collected data. 3) We utilize Partially Observable Markov Game (POMG) to disclose the game relation between multiple vehicles for the given task of trajectory planning so as to optimize conjoint deciders. 4) We use Non-dominated Sorting Genetic Algorithm (NSGA-II) as well as its improved versions to solve multiple-vehicles-based game dynamics model so as to find the optimal combinations of controlled variables of trajectory planning and communication. The research deliverables are expected to advance the theory and key technology of the state of the art trajectory planning method, the communication protocol design and the joint controlling for the future vehicles.

旨在以驾驶行为的安全要求和网络资源的总量限制为研究主线和优化约束，以轨迹规划和通信协作的控制量为研究内容和优化对象，以自适应情景的最大化轨迹规划性能和通信性能为研究目的和优化目标，以DOT、POMDP、POMG和NSGA-II为建模与计算方法，递进建立属性耦合关系图、单车情景感知博弈决策模型和多车博弈动力学模型并最终求解多目标、多约束优化解集。1)运用DOT建立覆盖情景感知、轨迹规划、通信协作和行为结果定性的属性耦合关系图；2)根据该图体现的控制量间和属性间不同耦合关系，运用POMDP建立单车情景感知博弈决策模型；3)根据该模型和多车间的合作与竞争关系，运用POMG建立多车博弈动力学模型；4)运用NSGA-II及其改进算法求解多车博弈动力学模型的优化解集。本项目定量地研究了智能网联汽车轨迹规划与通信协作的联合优化问题，为新轨迹规划方法新通信协作协议的设计以及联合控制提供理论依据与关键技术。

本项目以驾驶行为的安全要求和网络资源的总量限制为研究主线和优化约束，以轨迹规划和通信协作的控制量为研究内容和优化对象，以自适应情景的最大化轨迹规划性能和通信性能为研究目的和优化目标，以DOT、POMDP、POMG和NSGA-II为建模与计算方法，递进建立了属性耦合关系图、单车情景感知博弈决策模型和多车博弈动力学模型，并基于构建的模型求解多目标、多约束优化解集。首先，本项目运用DOT建立了覆盖情景感知、轨迹规划、通信协作和行为结果定性的属性耦合关系图；其次，基于图中所体现的控制量间和属性间不同耦合关系，运用POMDP建立了单车情景感知博弈决策模型；此外，本项目根据构建的模型和多车间的合作与竞争关系，运用POMG建立了多车博弈动力学模型；最后，本项目利用NSGA-II及其改进算法求解了多车博弈动力学模型的优化解集。本项目定量地研究了智能网联汽车轨迹规划与通信协作的联合优化问题，为新一代B5G通信中新轨迹规划方法新通信协作协议的设计以及联合控制提供了理论依据与关键技术。