Robust Navigation of Autonomous Transport Units in the Extractive Industry

Abstract:
Autonomous driving in temporarily GNSS-denied environments is challenging. Vehicle controllers of autonomous transport units require continuously precise information about vehicle position, speed, and heading. To cope with areas with no satellite signal reception, we introduce a multi-sensor navigation filter for articulated vehicles that fuses measurements from an inertial measurement unit (IMU), global navigation satellite systems (GNSS), wheel encoders, an optical speed sensor, and a barometer. Non-holonomic vehicle constraints are considered as well in the state estimation. The navigation filter uses two unscented Kalman filters (UKF) with a global fusion of the locally estimated states. This approach achieves improved robustness regarding single sensor failures compared to a centralized integration of all sensors in one filter. The developed navigation filter is evaluated experimentally with an articulated dumper in a gravel pit. With the proposed method, we achieved a mean position error of 0.19 m during a 190 s test drive in a gravel pit with a simulated GNSS interruption of 90 s.