不确定环境下基于V2X技术的多智能体协作定位研究

Currently the localization ability of unmanned autonomous autonomous driving vehicles represented by autonomous car is the key technology of navigation and mapping, environment understanding and decision-making and control. The measurements of sensors are easy to interfere and even lost under uncertain environment, such as big snow and big frog, unstructured road and electromagnetic interference, and jeopardize the localization performance of unmanned autonomous vehicles. Multi-sensor fusion by utlizing V2X communication techonology is one effective approach. Aiming at unknown system and measurement noise statistics, sensitiveness to outliers and high communication burden problems, this project study the distributed noise adaptive estimation algorithm, outlier detection and energy-efficient sensor scheduling methods, which increases the accuracy, robustness and communication ability. The proposed algorithms are planned to be tested and verified in autonomous car.

当前以自主驾驶车为代表的无人自主运动体的定位能力是实现其导航建图、环境理解和决策控制的关键。在大雪大雾、非结构化道路、电磁干扰等不确定环境下，传感器观测易受干扰甚至丢失，从而恶化无人自主运动体的定位能力。利用V2X通讯技术采集多传感器信息加以融合，实现自适应噪声抑制是当前一种有效方法。本项目针对动态不确定环境协作定位存在的系统和观测噪声特性未知、定位对异常值敏感以及通讯负担重的问题，分别重点研究分布式噪声自适应估计、异常值检测以及能效优先的传感器调度方法加以解决，从而提高协作定位的精度、鲁棒性和通讯能力。本项目进一步将所提方法在无人驾驶车上进行量化验证和测试。

当前车路协同系统中，自主驾驶的定位精度和对异常值的鲁棒能力是实现无人车安全可靠执行导航、路径规划、和控制任务的关键。然而由于运动模式的时变特性、传感器观测的不确定性、会存在系统建模误差未知、观测特性未知甚至观测异常以及通讯负担重等问题，从而恶化了定位的精度和鲁棒能力。针对这三个问题，本项目提出了面向动态不确定环境下多智能体协作定位的一体化技术框架。具体是在通过V2X通讯技术采集多平台多传感器的信息后，基于分布式自适应学习理论，分别重点研究系统噪声特性的多智能体自适应估计算法、异常值鲁棒的观测噪声特性的多智能体协同自适应估计算法以及能效优先的传感器调度方法加以解决，从而提高车路协同场景中协作定位的精度、鲁棒性和通讯能力。本项目拟在无人驾驶车路协同环境下进行量化验证和测试。

内容获取失败，请点击重试