A Hierarchical Graph-Based Terrain-Aware Autonomous Navigation Approach for Complementary Multimodal Ground-Aerial Exploration

Akash Patel, Mario A.V. Saucedo, Nikolaos Stathoulopoulos, Viswa Narayanan Sankaranarayanan · 2025

Read A Hierarchical Graph-Based Terrain-Aware Autonomous Navigation Approach for Complementary Multimodal Ground-Aerial Exploration by Akash Patel, Mario A.V. Saucedo, Nikolaos Stathoulopoulos, Viswa Narayanan Sankaranarayanan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-14859.txt.

Introduction

Akash Patel and co-authors guide a ground robot and a drone as one team through layered maps that judge terrain by risk and reward, in work from twenty twenty-five. Unknown spaces strain solo robots: wheels stall where wings would pass. Teamwork needs a shared sense of where each partner belongs.

Their framework stacks graphs at several resolutions, holding shape and meaning about passable ground. A confidence score weighs likely new volume, path safety, and collision risk. The ground robot uses it to decide when to send the drone ahead. In a real cave like trial, the ground robot flagged dead ends fit for flight. Shared frontier targets let the drone push past obstacles.

Catalog record

CREATORAkash Patel, Mario A.V. Saucedo, Nikolaos Stathoulopoulos, Viswa Narayanan Sankaranarayanan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-14859

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A Hierarchical Graph-Based Terrain-Aware Autonomous Navigation Approach for Complementary Multimodal Ground-Aerial Exploration Thanks: The authors are with Robotics & AI Team, Department of Computer, Electrical and Space Engineering, Luleå University of Technology, Luleå SE-97187, Sweden. Corresponding author: [email protected] Thanks: This work has been partially funded by the European Unions Horizon 2020 Research and Innovation Programme under the Grant Agreement No. 101138451 PERSEPHONE. Akash Patel, Mario A.V. Saucedo, Nikolaos Stathoulopoulos, Viswa Narayanan Sankaranarayanan Abstract Autonomous navigation in unknown environments is a fundamental challenge in robotics, particularly in coordinating ground and aerial robots to maximize exploration efficiency. This paper presents a novel approach that utilizes a hierarchical graph to represent the environment, encoding both geometric and semantic traversability. The framework enables the robots to compute a shared confidence metric, which helps the ground robot assess terrain and determine when deploying the aerial robot will extend exploration. The robot’s confidence in traversing a path is based on factors such as predicted volumetric gain, path traversability, and collision risk. A hierarchy of graphs is used to maintain an efficient representation of traversability and frontier information through multi-resolution maps. Evaluated in a real subterranean exploration scenario, the approach allows the ground robot to autonomously identify zones that are no longer traversable but suitable for aerial deployment. By leveraging this hierarchical structure, the ground robot can selectively share graph information on confidence-assessed frontier targets from parts of the scene, enabling the aerial robot to navigate beyond obstacles and continue exploration. I Introduction In the field of robotic exploration, ground and aerial robots have proven their pioneering capabilities to explore challenging terrains in underground deployment. Ground robots, such as wheeled robots and legged robots, are known for their extended operational endurance. On the other hand, aerial robots, such as drones are particularly interesting in 3D voids and vertical passages-like structures. The aerial robots have lower endurance limit compared to the ground robots; however, they can operate in situations where ground robot are simply unable to continue, for example, flying over obstacles. Inspired from the lessons learned during the DARPA subterranean challenge [ 1 ] [ 2 ] , this article aims at laying a foundation for a combined mobility robotic system in which the ground robot can identify deployment points using multimodal traversability information, encoded into navigation graphs. There exist multiple exploration architectures tailored for ground and aerial robots [ 3 ] [ 4 ] [ 5 ] deployed in the DARPA subterranean challenge. Our previous work on STAGE planner [ 6 ] focuses on graph based exploration approach tailored to solve dynamic uncertainty in the explored map while re-routing robot only using graph updates. Fig. 1: The complementary ground-aerial robotic system collaboratively explores a challenging subterranean environment. The ground robot autonomously identifies deployment points, based on multi-modal traversability and encoded within hierarchical graphs, enabling continued aerial exploration. In this article we leverage a graph based approach, as a baseline exploration framework, while incorporating a novel strategy that maintains hierarchy of graphs, encoding geometric and semantic traversability for the ground robot to assess exploration paths. The literature shows number of works attempting at ground-aerial robotic deployment in exploration scenarios. Lindqvist et al. in [ 7 ] presented a robotic system combining legged and aerial mobility for search and rescue operation. The legged robot carries the aerial robot to a manually defined deployment point, followed by releasing the aerial robot to explore using a reactive tunnel following method. This method particularly lacks in autonomy for a collaborative exploration mission, since the deployment point is manually defined by an operator and no knowledge is shared between the robots for cooperative exploration. De Petris et al. in [ 8 ] presented the work of team CERBERUS [ 9 ] in DARPA subterranean challenge work on marsupial walking-flying robotic exploration. In this method, when the legged robot reaches local exploration completion, it identifies partially seen areas that can be beneficial from aerial exploration. This approach also utilizes a graph based exploration planner [ 3 ] for the legged robot to maintain a graph composed of sparsely sampled vertices connected with robot’s poses and frontier vertices. The legged robot identifies frontier in blocked part of the tunnel and shares explored map with aerial robot to start exploration from the shared frontier. While this approach [ 8 ] presents a higher level of autonomy for the collaborative mission, it relies on sharing explored map and manually co-localizing the robots. Furthermore, the ground robot only relies on occupancy map information to depict the barrier as an untraversable area meaning the negative obstacles (steep stairs, or cavities in the ground) would not be identified as potential deployment zone for aerial robot. On the application side of the ground-aerial collaborative exploration, the work presented in [ 10 ] and [ 11 ] presents the missions for target identification and localization while utilizing navigation capabilities of heterogeneous robotic system. The related works on ground-aerial cooperative exploration studies shows that the heterogeneous robotic systems mostly rely on pre-defined scenarios to identify when aerial robot should be deployed and what level of information should be shared between the robots. In contrast to that, the proposed method incorporates multimodal traversability and frontier information to construct a robust confidence metric that helps the ground robot to autonomously identify deployment zone. On the other aspect of multi robot collaboration, the proposed approach allows the ground robot to have control over what level of minimal information needs to be shared with aerial robot for rapid deployment without requiring to share the full map. An example of fully autonomous ground-aerial collaborative exploration using the proposed approach is shown in Fig. 1 . II Contributions Building on the limitations of current state-of-the-art ground-aerial collaborative exploration methods, the proposed approach tackles key challenges and advances the autonomy of complementary robotic systems through the following novel contributions. The first contribution of this work introduces a novel approach that incorporates geometric and semantic traversability information to construct a hierarchical graph representing the robot’s understanding of the terrain. The approach integrates 3D LiDAR data into a high resolution grid structure to define geometric characteristics, such as elevation, slope, and surface roughness, within a 2.5D map. Additionally, it offers a fundamental improvement in voxelized representation within 3D occupancy maps by integrating semantic traversability values directly into the voxel state. Vertices along the exploration path are projected onto the 2.5D map to query geometric traversability, while ray casting is used on the 3D voxel map to query semantic traversability, allowing the robot to calculate its confidence in traversing a particular path. Based on this confidence metric the ground robot identifies untraversable zones where the ground-aerial robotic system could benefit from the aerial deployment to access the inaccessible zones. The second contribution introduces an ad hoc information-sharing mechanism between the ground and aerial robots for high-level knowledge transfer without the need to share explored maps. By utilizing a hierarchical graph for navigation, the ground robot can directly incorporate the frontiers that can benefit from aerial exploration into a shared graph. This allows for an ad-hoc dynamic graph sharing with the aerial robot without the need of sharing explored map. As part of the integration of autonomy modules for real life subterranean exploration,

Read the complete text