Visual Inertial Wheel Odometry on SE(3) for Mobile Robots with Extrinsic Calibration

The fusion of data from cameras, IMUs, and wheel encoders has demonstrated its effectiveness in mobile robot's localization. In this paper, we present a novel optimization-based Visual-Inertial-Wheel odometry to improve positioning accuracy. Specially, we model robot's pose with the Lie group SE(3) and derive the sensor measurement model on the corresponding Lie Algebra. Meanwhile, the proposed framework also allows to online calibrate the extrinsic parameters between the camera and IMU, the wheel encoder and IMU. Compared with other VIO methods based on point features, experimental results on KAIST's complex Urban Dataset show that our method has better performance on accuracy and robustness.