面向城市道路的多传感器融合定位导航技术

方便地获取高精度、高可靠的轨迹数据是交通、旅游等行业智能化发展的关键。鉴于此，本文设计一款集GPS、SINS和OBD于一体的多源车载组合导航系统来收集轨迹数据，提出数据融合的方法来弥补因传感器噪声导致的位置累积误差，并在GPS失锁时有效预测轨迹位置信息。此方法通过梯度提升与决策树相结合建立INS误差补偿模型，并引用粒子群算法优化模型的回归参数，可有效避免误差积累；再利用联邦滤波器实现GPS、SINS和OBD数据融合，提高了轨迹信息的准确性。实际道路测试证明，基于此方法的组合导航系统，在多种路况下可收集连续精准的轨迹数据。

Multi-sensor fusion for navigation technology and trajectory prediction under urban roads

Easy access to accurate and reliable trajectory data is the key to the intelligent development of transportation, tourism and other industries. In view of this, a multi-source vehicle integrated navigation system integrating GPS, SINS and OBD is designed to collect trajectory data. The FGP data fusion method is proposed to compensate the accumulated position error caused by sensor noise, and to predict the track position information when the GPS outage. To effectively avoid error accumulation, the INS error compensation model is established through combination of Gradient Boosting and Decision Tree, and the regression parameters of the model are optimized by Particle Swarm Optimization in this method. The GPS, SINS and OBD data fusion is realized by exploiting federal filter thereby improving the accuracy of trajectory information. The actual road test proves that the GSO device based on the FGP can collect continuous and accurate trajectory data under various road conditions.