多目标环境下车辆自主决策可达集分析

由于自主车辆具有离散控制和连续行为混成以及不可避免的系统偏差存在的特点，使传统验证方法无法遍历所有不确定的系统输入和运行状态，因而自主车辆安全决策的置信水平较低。为此，提出以车道边界作为约束条件的多目标约束的可达集建模方法，该方法利用哈密顿雅可比方程控制可达集变化，利用水平集完成可达集的可视化任务，确定具有不确定系统输入和运行状态的自主车辆安全区域。结果表明，与传统仿真方法相比，基于可达集的自主车辆多目标安全决策方法具有更高的适应性,并且可以为智能驾驶车辆在较为复杂环境下的安全性提供理论保证。

Reachability Set Analysis of Vehicle Autonomous Decision in Multi-objective Environment

Due to the autonomous vehicles have characteristics of discrete control, continuous mixed behavior and inevitable system deviation, the traditional method cannot traverse all uncertain system import and operation states. Therefore, the confidence level of safety decision is low. This paper proposed a reachable set model with multi-objective constraints which set the lane boundaries as constraints.This method uses Hamiltonian Jacobi equation to control the change of reachability set and uses level set to accomplish the target of visualization with reachable set. Besides, it determines the autonomous vehicle safety area with uncertain system import and operation states. The results show that the multi-objective safety decision method based on reachability set has higher adaptability than traditional simulation methods. Meanwhile, it can provide theoretical guarantee for the safety of intelligent car in complex environments.