OpenStreetMap-Based Autonomous Navigation With LiDAR Naive-Valley-Path Obstacle Avoidance

OpenStreetMaps (OSM) is currently studied as the environment representation for autonomous navigation. It provides advantages such global consistency, a heavy-less map construction process, and wide variety of road information publicly available. However, location this usually not very accurate locally. In paper, we present complete navigation pipeline using OSM planning. To avoid flaw local low-accuracy, offer novel LiDAR-based Naive-Valley-Path (NVP) method that exploits concept "valley" areas to infer path always furthest from obstacles. This behavior allows through center trafficable following road's shape independently error. Furthermore, NVP naive highly sample-time-efficient. time efficiency also enables obstacle avoidance, even dynamic objects. We demonstrate system's robustness in our research platform BLUE, driving autonomously across University Alicante Scientific Park more than 20 km with 0.24 meters average error against 19.8 ms sample time. Our vehicle avoids static obstacles ones, vehicles pedestrians.