面向无人驾驶主动安全保障的有轨电车状态估计与风险评估

Mastering the real time state of tram and assessing the potential dangerous situation (e.g. collision) in the surrounding environment are key issues to ensure unmanned tram running safety. However, because of the nonlinear feature and parameter drift of the dynamic model of vehicles, the interference of stochastic noise, and its special running environment, it is hard to master the running state in real time, and difficult to assess the hazard of running safety from the obstacles such as pedestrians and cars around. Regarding the problems above, this research plans to carry out the research including the following three aspects. Firstly, a set of parametric state space models is established based on the dynamic characteristics of tram and online measuring information, and a nonlinear parameter identification algorithm and method is put forward to obtain the actual dynamic model at different running condition; Secondly, after an investigation of the transition conditions, a nonlinear interacting multiple model is established to characterize the running process of tram, and a state estimation algorithm for nonlinear and non-Gaussian interacting multiple model (NNIMM) is proposed to grasp the tram running state; Lastly, after a study of active and passive safety protection mechanism, a safety assessment method and model is proposed regarding the potential risk of collision, and a comprehensive objective optimization model is established by considering safety, energy consumption, punctuality and comfortability, to obtain the optimal running strategy. The research results can provide theoretical basis and technical guidance for the unmanned tram system design and development. Also, it is of great significance for promoting and guaranteeing safety of the unmanned tram system which has independent intellectual property rights of our country.

对列车自身运行状态的把握,以及对潜在的碰撞等周围环境中危险状况的准确评估是保障无人驾驶有轨电车运行安全的关键。然而，由于车辆动力学模型的非线性特征和参数漂移、随机噪声干扰及运行环境复杂多变，致使列车的实际运行状态往往难以估计；同时，周围行人、汽车等其他路权者对列车安全运行的影响也难以评估。针对以上问题，本项目主要展开以下三方面的研究：基于列车动力学特性和在线观测信息集，建立典型运行工况下的状态空间参数模型集合，提出参数辨识理论与方法，以获取各工况下准确的动力学模型；研究各工况的转换机理，构建列车运行的非线性交互多模型，提出状态估计理论与方法，以获取列车的实际运行状态；对外部潜在风险进行量化评估，建立列车主动安全评估方法，构建基于安全、节能、准点、舒适度以及精确停车的多目标优化模型，以探索综合优化行车策略。研究成果可为开发具有我国自主知识产权的有轨电车无人驾驶系统提供理论基础和技术支撑。

针对无人驾驶有轨电车主动安全保障的关键问题，本课题研究了无人驾驶有轨电车的运行状态估计方法，以及基于数据驱动的无人驾驶风险评估方法。首先,针对各运行工况下的状态空间参数模型，项目组研究分析了列车运行的动力学特性, 构建了列车离散化制动模型。同时考虑到列车速度和制动盘摩擦系数的时间序列数据特性，提出了时变制动参数辨识方法，实现利用列车制动速度的历史数据来预测未来的制动盘摩擦系数变化。在此基础上，项目组研究了非线性状态空间模型的在线状态估计和参数辨识理论与方法，获取有轨电车各运行工况下的准确动力学模型，实现了在任意随机干扰下，列车非线性动力学模型的参数辨识，为有轨电车的实时运行状态估计奠定模型基础。其次，针对列车运行过程中动力学模型的交互性，项目组研究了列车运行中各工况的变化所引起的动力学模型转换的条件和规律，建立基于工况变换的非线性交互多模型。同时，项目组提出了计算车辆碰撞概率的方法，建立了有轨电车运行状态的估计理论与方法，实现列车对自身状态、速度-制动距离关系的实时、准确掌握。最后，针对无人驾驶的有轨电车安全保障机理，量化周围环境对列车运行安全造成的风险，项目组基于上述获得的列车运行状态，量化各影响因素对列车安全运行的风险程度，建立对应风险评估函数，完成基于风险评估的列车运行安全评估建模与分析，为实现有轨电车的高安全、高舒适度、高准点率、精确位置停车提供理论基础。通过以上研究，形成了从状态估计到风险评估的一系列主动安全保障理论与方法，为无人驾驶有轨电车的安全运行奠定了理论基础。