Robust Vehicle State Estimation From Urban Driving to the Limits of Handling

High-precision vehicle position, velocity, and attitude estimation on any terrain with full-range vehicle motion is the next step to improve the availability of advanced vehicle safety and autonomous systems. Estimating vehicle motion from a very slow speed to beyond the limits of handling under various conditions not only on asphalt but also on snow with sufficient accuracy using automotive level sensors is a significant challenge. This paper presents a robust vehicle state estimation algorithm that covers full-range vehicle motion under various road and maneuvering conditions. An inertial measurement unit (IMU) is used as the main sensor, whereas an adaptive vehicle model is used as a stochastic constraint to enhance the accuracy and resilience of the algorithm. Furthermore, unknown sensor delays are estimated for sensors that are not time-synchronized, thereby resulting in a zero-delay vehicle state output. Four real car experiments were conducted: urban driving, racetrack sports driving, drift driving, and snow/ice driving with large sliding motions. The position, velocity, and attitude errors were less than 0.1 m, 0.1 m/s, and 0.1 deg, respectively, under all conditions. Furthermore, the position errors during 8 and 96 s of dead-reckoning were below 0.1 and 0.66 m, respectively. The proposed algorithm is expected to enhance the safety and improve the availability of autonomous vehicle safety systems.