Abstract
In this paper, we address the current state of the need for accurate real-time positioning of vehicles in urban environments and the low precision of existing Kalman-based in-vehicle GNSS navigation algorithms, considering the fusion of navigation positioning and visual environment perception. we propose a new multi-mode adaptive in-vehicle GNSS navigation and positioning algorithm based on environment perception in urban environments. First, we use low-cost environment perception devices to capture vehicle environment information in real time and determine the vehicle driving state based on vehicle status switching and the optical flow method. Second, based on the already determined vehicle driving state, we adapt the vehicle state model to solve the problem of low precision of existing Kalman-based GNSS vehicle navigation and positioning algorithms. Then, considering that vehicles frequently stop during operation, we accurately determine the vehicle's location information based on the judgment of the vehicle's stationary state. Finally, after verifying the effectiveness and reasonableness of the algorithm through simulation experiments, we design actual in-vehicle navigation experiments, and compared with the common algorithm, our proposed algorithm significantly improves the actual GNSS positioning accuracy of vehicles in urban environments by 27%.