基于隐马尔可夫链模型的车道级地图匹配
Lane-Scale Map Matching Based on Hidden Markov Model

稳定的车道级地图匹配对于自动驾驶系统至关重要。地图匹配技术一直是车载融合定位与高精地图语义信息的重要连接。本文提出了一种基于隐马尔可夫模型的实时车道级地图匹配方法，将车路融合感知计算下的车辆轨迹与车道级道路交通地图，进行匹配计算。在构建基于车道级地图的状态转移概率时，本文考虑了车道及其拓扑关联、车道线类型及换道许可规则、定位误差模式，并构建了隐马尔科夫模型中发射和转移概率等参数的数理模型。实验证明，本模型在不同的融合感知定位精度水平下，实现了较高的匹配精度。在主动添加1.5 m随机误差的条件下，依旧实现了96.09%的召回率和0.983 m的位置偏差。
Stable lane-level map matching is critical for autonomous driving systems. Map matching technology has always been an important connection between vehicle fusion positioning and high-precision map semantic information. This paper proposes a real-time lane-level map matching method based on the hidden Markov model, which matches the vehicle trajectory under the vehicle-road fusion perception calculation with the lane-level road traffic map. When constructing the state transition probability based on the lane-level map, this paper considers the lane and its topological association, lane line type and lane change permission rules, and the positioning error mode, and constructs the mathematical model of parameters such as launch and transition probability in hidden Markov model. Experiments have proved that this model achieves high matching accuracy at different levels of fusion perception positioning accuracy. Under the condition of actively adding 1.5 m random error, the recall rate of 96.09% and the position deviation of 0.983 m are still achieved.