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

TL;DR

This work addresses local trajectory deviations caused by inaccuracies in OpenStreetMap when used for global planning by introducing a LiDAR-based Naive-Valley-Path (NVP) for Local Path Planning within a complete OSM-based autonomous navigation pipeline. The approach leverages valley-cost maps and a naive, concentric-circle inference to steer the vehicle along the center of trafficable regions, enabling real-time obstacle avoidance including dynamic objects. The system, demonstrated on the BLUE platform over extensive outdoor tests, achieves a mean lateral error of about $0.24$ m from the road center and maintains robust performance under static and dynamic obstacles, outperforming a contemporary OSM-based method in both path fidelity and responsiveness. The combination of global OSM graph-based planning with fast NVP-based local planning offers globally consistent navigation with locally accurate, center-following trajectories, suitable for unstructured outdoor environments. Future work explores dynamic graph updates via JOSM and improved perception/localization integration to further enhance robustness and scalability.

Abstract

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

Figures (13)

• Figure 1: The complete autonomous navigation pipeline, where path planning approach is organized hierarchically. First, the GPP gets information about global localization from the robot and global information about the environment from OSM to plan a global path. Then, the LPP module, using the presented Naive-Valley-Path (NVP) method, recalculates the local path to obtain the optimal way to follow.
• Figure 2: Example of road network (graph) used in the GPP module, in this case, extracted from the Scientific Park area in the University of Alicante. The red points describe the georeferenced position of nodes, while the blue lines represent the links that indicate a trafficable connection between nodes. We can extract the graph directly from current OSM data or create it manually in the JOSM application.
• Figure 3: Different top-view 2D representations of LiDAR information for Valley-Path calculation given a target goal represented as a red star: a) Projection in a top-view 2D plane of obstacles point cloud $\mathcal{P}^o$. b) Free-space map, where the blue area represents the space free of obstacles. c) Cost map defined in (2). d) Inverted representation of the gradient magnitude of c), which shows clear possible paths in the valley areas.
• Figure 4: The Naive-Valley-Path (NVP) Calculation. The red points are the ones that form the local path $\mathbf{P}^l = \left(\mathbf{p}^{c_0}_{nn}, \mathbf{p}^{c_1}_{nn}, ..., \mathbf{p}^{c_N}_{nn} \right)$. The green connections describe the naive assumption and define the angle of the points. For the sake of clarity, this representation shows the NVP superpose with non-naive representation.
• Figure 5: The arcs represent the possible trajectories of control actions $\mathbf{u}_i$. The red ones are the collision-risk trajectories, and their corresponding control actions are discarded. In contrast, the green ones are collision-free.