A Fuzzy-Innovation-Based Adaptive Kalman Filter for Enhanced Vehicle Positioning in Dense Urban Environments

In this paper, a fuzzy-innovation based adaptive extended Kalman filter (FI-AKF) is proposed to improve the performance of the GNSS/INS fusion system, which is degraded due to satellite signal cutoff and attenuation and inaccurate modeling in dense urban environments. The information used for sensor fusion is obtained from real-time kinematic (RTK), micro-electro-mechanical system based inertial measumrement unit (MEMS-IMU), and on-board diagnostics (OBD). The fuzzy logic system is proposed to adaptively update the measurement covariance matrix of the RTK according to the position dilution of precision (PDOP), the number of receivable satellites, and the innovation of the extended Kalman filter (EKF). In addition, the driving state of the vehicle is defined as stop, straight run, left/right turn, and the like. To reduce the heading estimation error of the Kalman filter, the estimated heading is corrected according to the driving state. Also, the measurement covariance matrices of IMU and OBD are applied adaptively considering the characteristics of each sensor according to the driving state. In order to analyze the performance of the proposed FI-AKF positioning system in a dense urban environment, a computer simulation is performed. The proposed FI-AKF is compared to the performance of the existing extended Kalman filter and the innovation-based adaptive extended Kalman filter. In addition, we conduct a performance comparison experiment with a commercial positioning system in the field test. Through each experiment, it is confirmed that the proposed FI-AKF system has higher positioning performance than the comparison positioning systems in a dense urban environment.