DART: Dual-level Autonomous Robotic Topology for Efficient Exploration in Unknown Environments

TL;DR

The paper addresses inefficiencies in autonomous exploration by introducing a dual-level topological framework that uses a Low-level Topological Graph (LTG) for rapid frontier detection and a High-level Topological Graph (HTG) for non-greedy region prioritization, complemented by a Local Artificial Potential Field (LAPF) for fast motion control. The LTG and HTG are generated through uniform sampling, graph erosion, and trail-marker connections, enabling systematic exploration of unfinished convex regions while avoiding impassable routes. Experimental results in simulation and real-world tests show substantial improvements in exploration speed and travel distance, with ablation studies confirming the contribution of each component and robustness to mapping inaccuracies. The approach offers a practical, scalable solution for efficient exploration in structured, partially known environments and opens avenues for extension to dynamic or semantically enriched maps.

Abstract

Conventional algorithms in autonomous exploration face challenges due to their inability to accurately and efficiently identify the spatial distribution of convex regions in the real-time map. These methods often prioritize navigation toward the nearest or information-rich frontiers -- the boundaries between known and unknown areas -- resulting in incomplete convex region exploration and requiring excessive backtracking to revisit these missed areas. To address these limitations, this paper introduces an innovative dual-level topological analysis approach. First, we introduce a Low-level Topological Graph (LTG), generated through uniform sampling of the original map data, which captures essential geometric and connectivity details. Next, the LTG is transformed into a High-level Topological Graph (HTG), representing the spatial layout and exploration completeness of convex regions, prioritizing the exploration of convex regions that are not fully explored and minimizing unnecessary backtracking. Finally, an novel Local Artificial Potential Field (LAPF) method is employed for motion control, replacing conventional path planning and boosting overall efficiency. Experimental results highlight the effectiveness of our approach. Simulation tests reveal that our framework significantly reduces exploration time and travel distance, outperforming existing methods in both speed and efficiency. Ablation studies confirm the critical role of each framework component. Real-world tests demonstrate the robustness of our method in environments with poor mapping quality, surpassing other approaches in adaptability to mapping inaccuracies and inaccessible areas.

Figures (9)

• Figure 1: Illustration of the dual-level topological exploration Graph. (a) The LTG, generated by uniformly sampling the original map, serves as the foundation for subsequent exploration algorithms. (b) The HTG, derived from processing the LTG, shows not only the distribution and connectivity of the convex polygon regions but also whether each region has been fully explored.
• Figure 2: The dual-level topological graph framework comprises three main components. The first component, the LTG module, is constructed through uniform sampling of the occupancy grid map to extract frontier information. The second component, the HTG module, integrates trail marker nodes with region nodes derived by eroding the LTG, facilitating structured exploration planning. The final component, the motion control module, employs a LAPF method to directly command the robot's movements, optimizing navigation based on real-time data and graph information.
• Figure 3: Illustration of graph frontier structuring: (a) The occupancy grid map. (b) Preliminary filtering of peripheral nodes with degrees less than 8 for frontier nodes extraction. (c) Diffusion from previously filtered nodes, expanding until reaching occupied cells or diffusion depth limit. (d) BFS-based clustering of frontier nodes for major area identification.
• Figure 4: The process of graph erosion segmentation: (a) The erosion on LTG. The color gradient from lighter blue to darker red indicates the sequence of erosion, with lighter colors representing nodes eroded earlier. (b) The generation of region nodes. All region nodes, generated through the erosion process, are matched with frontier clusters to determine whether the region has been fully explored.
• Figure 5: Illustration of target frontier cluster selection and LAPF: (a) Our approach focuses on exploring incompletely explored convex polygon regions, unlike traditional methods that prioritize high-entropy clusters, leading to reduced backtracking. (b) LAPF evaluates several initial nodes from the path nodes to integrate attraction and repulsion forces, determining the robot's optimal movement direction.