In short:
To ensure that our experimental vehicle can reliably and accurately localize itself in its global reference frame, we have designed a geolocation system based on multi-sensor data fusion using an Extended Kalman Filter. This system leverages data from two GNSS antennas, a 6-axis inertial measurement unit (IMU), and wheel encoders. During development, we conducted a comprehensive theoretical study to validate the solution in simulation before integrating it into our vehicle. These findings were published at the AgriControl conference in August 2022.
Abstract:
This paper focuses on the development of a data fusion architecture for positioning and attitude estimation of an autonomous agricultural vehicle, combining two RTK-GNSS and an Inertial Measurement Unit (IMU) system. The main algorithm steps are presented giving a generic approach for real-time vehicle guidance applications or localization using data fusion. Important features such as sensor error modeling based on the Allan variance method as well as compensation phenomena related to terrestrial navigation using IMU mechanization are presented. A loosely coupled fusion architecture is proposed allowing low complexity for real-time algorithm integration. Finally, results based on real data from a real prototype are exploited to show the efficiency of the proposed algorithm.