Multi-modal odometry estimation via EKF-based feedback architecture

Accurate odometry is fundamental for most mobile robot applications. In this paper, we propose a multimodal sensor fusion architecture for odometry estimation based on an Extended Kalman Filter (EKF), which integrates asynchronous data sources such as wheel odometry, IMU measurements, and LiDAR-based odometry estimates to exploit the strength of each sensor. To correct drift, we employ LiDAR odometry using an algorithm (Kinematic ICP) that refines estimates using point cloud alignment constrained by robot's kinematics. Its output is fed back into the EKF, forming a closed-loop correction mechanism. We evaluate the system's performance in a simulated environment with real-world conditions.