Pose Estimation for Spherical Robot with Kinematics-Based Encoder-Inertial Odometry

The spherical robot based on pendulum is a novel type of mobile robot whose motion is realized by changing the center of mass. Since the exteroceptive sensors such as camera and LiDAR are significantly hindered or even uninformative inside the spherical shell, the pose estimation for spherical robot is challenging. In this paper, we present a concise kinematic model for pendulum-driven spherical robot and propose a low-cost kinematics-based encoder-inertial odometry algorithm to estimate position and orientation. Practical experiments results demonstrate that our approach is efficient for short-term localization. Despite the accumulated errors problem in long-term localization, our approach provides a baseline in further study on localization for pendulum-driven spherical robot and can serve as an initial estimate or interpolating instances for further simultaneous localization and mapping (SLAM) for spherical robots.