Vehicle Heading Estimation Using Positioning and Inertial Data-Based Adaptive Tandem Kalman Filter

This paper proposes an Adaptive Tandem Kalman Filter (ATKF) algorithmic method for accurate, robust, and magnetometer-free heading estimation through the competitive fusion of inertial and positioning data. It aims to mitigate the angular heading drift, typically present in MEMS IMU (Micro-Electromechanical Systems Inertial Measurement Unit) sensors, without the use of environmentally sensitive magnetometers. Due to its adaptive nature, the ATKF algorithm is capable of iterative estimation of the input data significance, based on the tracked vehicle behavior, and performing a corresponding weighted fusion. A novel tandem structure allows to perform a series of consecutive data processing steps within a single Kalman filter iteration, thus minimizing the delay in algorithmic response to the input data. Series of simulated comparison tests with one of the state-of-the-art algorithms have demonstrated the high stability and robustness of the proposed algorithm. It has shown a consistent 40% to 90% improvement in the estimated heading accuracy and precision, depending on the maneuvering intensity and the data quality. Simulation results were experimentally validated during the full-scale test campaign, conducted in the industrial environment using a highly maneuverable forklift. Real-time forklift heading was tracked by the proposed ATKF algorithm with a 1$^\circ$overall median error and 2.3$^\circ$median error during the active movement periods. The proposed method has respectively shown 95% and 93% improvement in initial IMU heading accuracy and precision. Experimental performance evaluation of the magnetometer-based heading estimation methods has practically confirmed their unreliability and inconsistency in industrial applications. The proposed ATKF heading estimation algorithm may find a variety of possible applications in the field of robotics and intelligent vehicles. It is expected to be especially useful in magnetometer-denied environments.