Contribution

Investigating 2.5D path-planning methods for autonomous mobile robots in complex unstructured off-road scenarios

---

Abstract

Most of the existing literature focuses on path planning in 2D, where the 3D world is converted to a 2D grid map. There is little literature on methods that can natively utilize 2.5D or 3D information and thus use a less compressed representation of the environment for planning. In this work, methods from both groups were systematically compared. A suitable simulator and physics engine have been chosen to enable a realistic evaluation of 2.5D navigation in a simulation. For the methods using the 2D view, classical and widely used planning algorithms were used. To generate the map for the classical methods, a 2.5D map was converted into a 2D map using slope information. The classical search algorithms find a path based on costs on the 2D map. To test a method that uses native 2.5D data for planning, a novel approach was developed that uses the robot’s orientations on a 2.5D elevation map.This method samples different locations on the 2.5D map and considers the attitude of the footprint for each position to generate the cost. The evaluation showed that the proposed method, which uses 2.5D data directly, planned shorter and faster paths in most scenarios, while the journey remained safe and reliable for the robot. The results for the classical, 2D methods showed that they are especially useful in scenarios where low computational power is available

Keywords: Path planning, Autonomous robot, rapidly- exploring random tree

How to Cite:

Koczka, A. & Steinbauer-Wagner, G., (2025) “Investigating 2.5D path-planning methods for autonomous mobile robots in complex unstructured off-road scenarios”, ARW Proceedings 25(1), 91-96. doi: https://doi.org/10.34749/3061-0710.2025.15