A comparative experimental study of LiDAR, camera, and LiDAR-camera localization algorithms for autonomous mobile robots on urban sidewalks

To navigate autonomously, mobile robots require an accurate and drift-free source of localization. Many algorithmic solutions have been proposed in the literature, exploiting sensors like LiDARs and cameras. This paper proposes a benchmark of some open-source localization algorithms on a self-acquired, sidewalk-focused dataset. The dataset was collected by an instrumented robotic platform traveling more than 5 kilometers on public sidewalks. The analysis targets the use case of map-based localization exploiting an a-priori known localization map. Results are reported as mean position errors with respect to a high quality ground truth trajectory, as computed by the Absolute Trajectory Error (ATE) method.