Online IMU-Odometer Extrinsic Calibration Based on Visual-Inertial-Odometer Fusion for Ground Vehicles

Aiming at the problem of autonomous positioning for the ground vehicles in the urban areas, we propose an efficient and practical vehicle positioning method using the measurements from the camera, IMU and odometer with online extrinsic calibration. A two-stage method is constructed to obtain the complete 5 degree-of-freedom IMU-odometer extrinsic parameters. Firstly, the roll and pitch of the orientation are estimated by static gravity direction. Secondly, the yaw and translation are estimated online based on the IMU-odometer residual in the back-end optimization, which is constructed to fuse the odometer measurements by non-holonomic constraint and level-arm effect compensation. The proposed method has been validated extensively with respect to the robustness and accuracy in simulation environment and autonomous driving scenarios of the real world.