Maximum Correntropy Kalman Filter for Orientation Estimation with Application to Lidar Inertial Odometry

Accurate attitude estimation using low-cost sensors is animportant capability to enable many robotic applications. Inthis paper, we present a method based on the concept of corren-tropy in Kalman filtering to estimate the 3D orientation of a rigidbody using a low-cost inertial measurement unit (IMU). We thenleverage the proposed attitude estimation framework to developa LiDAR-Intertial Odometry (LIO) demonstrating improved lo-calization accuracy with respect to traditional methods. Thisis of particular importance when the robot undergoes high-ratemotions that typically exacerbate the issues associated with low-cost sensors. The proposed orientation estimation approach isfirst validated using the data coming from a low-cost IMU sen-sor. We further demonstrate the performance of the proposedLIO solution in a simulated robotic cave exploration scenario