混行交通中时变非高斯噪声和定位失效下智能网联汽车目标跟踪

在以智能网联汽车为引领的“新基建”浪潮中，实时和准确的目标跟踪是打通智能网联汽车进入人们日常生活“最后一公里”的关键挑战之一。本项目针对智能网联汽车在混行交通中目标跟踪面临的“一变两未知”问题展开研究。具体而言，首先，针对单车单传感器目标跟踪，探索时变非高斯噪声和车体坐标系移动未知下的跟踪建模机理，提出基于自适应时变混合高斯和一致目标漂移的跟踪模型；然后，针对单车多传感器目标跟踪，提出基于变分贝叶斯的融合跟踪算法，突破噪声辨识、量测融合、数据关联、状态估计和车体坐标系移动估计的一体化求解难题，并构建定位失效下的航迹管理策略；在此基础上，针对多车协同跟踪，建立多车相对位置估计、航迹集成关联和融合的联合优化方法，攻克多智能网联汽车相对位置未知下的多航迹融合难题；最后，利用前期开发的仿真平台和真车实验，对所提跟踪算法进行验证，以期丰富智能网联汽车目标跟踪理论，为其规模化发展提供理论和技术支撑。

As for the intense need for connected and intelligent vehicles (CIVs), accurate and reliable multi-target tracking in the real-time manner is a critical issue for CIVs. This project focuses on the problem of multi-target tracking for CIVs with time-varying non-Gaussian noise and failure localization in mixed traffic scenario. Specifically, the methods of time-varying Gaussian mixture model and coherence target drift will be proposed to perform an unknown moving single-sensor multi-target tracking modelling with time-varying non-Gaussian noise. Moreover, a novel unknown moving multi-sensor multi-target tracking algorithm will be developed using a variational Bayesian technique. In this algorithm, outlier detection, measurement fusion, data association, state estimation, and self-localization for CIV are performed simultaneously. Then, a novel method of track management for failure localization is proposed. Furthermore, in order to perform the problem of cooperative tracking for multiple CIVs, a unified cooperative positioning, multiple tracks association and fusion algorithm is also to be established. Finally, the effectiveness of the proposed algorithms will be verified by simulations and experiments. The main purpose of this proposal is to break through key techniques concerned with multi-target tracking for CIVs with time-varying non-Gaussian noise and failure localization. The resulting theories and techniques will provide strong support for the large-scale development of CIVs.