Comparison of Trajectory Estimation Methods Based on LIDAR and Monocular Camera in a Simulated Environment

Odometry estimation is a fundamental problem for two-dimensional locomotion in mobile robotics. When common motion sensors, such as wheel encoders, are absent or unreliable, other sensors can be employed for the same task. LIDARs are a common choice due to their precision but have relatively high cost. The monocular camera provides a cost-effective alternative but has limitations, such as reliance on lighting conditions and the absence of direct depth information. Both sensors face specific challenges when employed in indoor environments with SLAM algorithms for odometry estimation. This article proposes a comparative analysis between the Hector SLAM algorithm, based on LIDAR, and the ORB-SLAM3 algorithm, based on a monocular camera, to assess the accuracy of estimated trajectories. The sensors were mounted on a wheeled robot in a simulated environment and the simulator provided the ground truth trajectories. As expected, it was observed that the LIDAR-based algorithm performed better than the camera-based one, but the latter is an acceptable replacement for odometry estimation when the trajectories are simple and executed at lower speeds.