Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Sensor-based Multi-agent Coverage Control with Spatial Separation in Unstructured Environments

Xinyi Wang, Jiwen Xu, Chuanxiang Gao, Yizhou Chen, Jihan Zhang, Chenggang Wang, Ben M. Chen

TL;DR

This work tackles robust, efficient coverage by a team of robots in unknown unstructured 3D environments using a decentralized Voronoi-based policy that operates directly on point-cloud data. It introduces spatial decomposition with robot-robot and robot-obstacle collision hyperplanes, and a safe region construction via a low-dimensional QP, ensuring collision avoidance and non-entrapment. A deadlock-aware guided map, powered by a navigation function computed with A*, steers exploration toward information-rich regions while avoiding local minima, and is integrated with a centroid-based CVT objective. Extensive high-fidelity simulations in forest-like and indoor environments demonstrate superior coverage ratio, higher success rates, and shorter task times compared with state-of-the-art Voronoi-based methods, highlighting practical viability for real-world deployments in complex settings where full maps are unavailable.

Abstract

Multi-robot systems have increasingly become instrumental in tackling search and coverage problems. However, the challenge of optimizing task efficiency without compromising task success still persists, particularly in expansive, unstructured environments with dense obstacles. This paper presents an innovative, decentralized Voronoi-based approach for search and coverage to reactively navigate these complexities while maintaining safety. This approach leverages the active sensing capabilities of multi-robot systems to supplement GIS (Geographic Information System), offering a more comprehensive and real-time understanding of the environment. Based on point cloud data, which is inherently non-convex and unstructured, this method efficiently generates collision-free Voronoi regions using only local sensing information through spatial decomposition and spherical mirroring techniques. Then, deadlock-aware guided map integrated with a gradient-optimized, centroid Voronoi-based coverage control policy, is constructed to improve efficiency by avoiding exhaustive searches and local sensing pitfalls. The effectiveness of our algorithm has been validated through extensive numerical simulations in high-fidelity environments, demonstrating significant improvements in both task success rate, coverage ratio, and task execution time compared with others.

Sensor-based Multi-agent Coverage Control with Spatial Separation in Unstructured Environments

TL;DR

This work tackles robust, efficient coverage by a team of robots in unknown unstructured 3D environments using a decentralized Voronoi-based policy that operates directly on point-cloud data. It introduces spatial decomposition with robot-robot and robot-obstacle collision hyperplanes, and a safe region construction via a low-dimensional QP, ensuring collision avoidance and non-entrapment. A deadlock-aware guided map, powered by a navigation function computed with A*, steers exploration toward information-rich regions while avoiding local minima, and is integrated with a centroid-based CVT objective. Extensive high-fidelity simulations in forest-like and indoor environments demonstrate superior coverage ratio, higher success rates, and shorter task times compared with state-of-the-art Voronoi-based methods, highlighting practical viability for real-world deployments in complex settings where full maps are unavailable.

Abstract

Multi-robot systems have increasingly become instrumental in tackling search and coverage problems. However, the challenge of optimizing task efficiency without compromising task success still persists, particularly in expansive, unstructured environments with dense obstacles. This paper presents an innovative, decentralized Voronoi-based approach for search and coverage to reactively navigate these complexities while maintaining safety. This approach leverages the active sensing capabilities of multi-robot systems to supplement GIS (Geographic Information System), offering a more comprehensive and real-time understanding of the environment. Based on point cloud data, which is inherently non-convex and unstructured, this method efficiently generates collision-free Voronoi regions using only local sensing information through spatial decomposition and spherical mirroring techniques. Then, deadlock-aware guided map integrated with a gradient-optimized, centroid Voronoi-based coverage control policy, is constructed to improve efficiency by avoiding exhaustive searches and local sensing pitfalls. The effectiveness of our algorithm has been validated through extensive numerical simulations in high-fidelity environments, demonstrating significant improvements in both task success rate, coverage ratio, and task execution time compared with others.
Paper Structure (15 sections, 4 theorems, 8 equations, 8 figures, 1 table)

This paper contains 15 sections, 4 theorems, 8 equations, 8 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Sensor and Obstacle Representation
  5. Voronoi-based Coverage Control
  6. Sensor-based Coverage Control with Deadlocks and Collision Avoidance
  7. Spatial Decomposition
  8. Robot-robot Collision Avoidance Hyperplane
  9. Robot-obstacle Collision Avoidance Hyperplane
  10. Safe Region Construction
  11. Deadlock-aware Guided Map
  12. Simulations
  13. High-fidelity Validation
  14. Comparisons of Coverage Performance
  15. Conclusions

Key Result

Lemma 1

If $\Vert \mathbf{p}_{i}(t) - \mathbf{p}_{j}(t) \Vert\geq r_i+r_j, i \neq j$ and $\Vert \mathbf{p}_i(t) -\mathbf{q}_o \Vert \geq r_i$, $\forall i,j \in 1,\dots,n, o\in 1,\dots,m_c$ at time $t$, we have 1) $\Bar{\mathcal{V}_i} \neq \emptyset$; 2)$\Bar{\mathcal{V}_i} \subset \mathcal{V}_i$; 3) $\Vert

Figures (8)

  • Figure 1: GIS provides maps as prior information for multi-robot systems in post-disaster search and rescue.
  • Figure 2: Multi-robot coverage mission in unstructured environments with three sparsely distributed information sources.
  • Figure 3: Illustration of spatial decomposition and safe region construction in environments filled with a high density of obstacles, as represented by point clouds. The original squares within sensor range $R_\mathrm{sensor}$ are projected outside of the circle with a one-to-one mapping. The robot is positioned in an obstacle-free location $\mathbf{p}_i$. By finding the convex hull of the mirrored points, i.e., $\mathbf{q}'_1,\dots,\mathbf{q}'_5$ (the yellow squares), we can target which squares that are closest to the robot, i.e., $\mathbf{q}_1,\dots,\mathbf{q}_5$ (the green squares). Subsequently, a safe convex region $\Bar{\mathcal{V}}_i$ colored in green can be obtained by separating hyperplane theorem.
  • Figure 4: Coverage of ToI in two 3D large-scale scenarios by 16 robots. (a) Cluttered and thin structural forest environment. (b) Narrow indoor office environment.
  • Figure 5: One trial performance in terms of computational time and minimum distance during coverage by 16 robots.
  • ...and 3 more figures

Theorems & Definitions (5)

  • Definition 1
  • Lemma 1: Properties of BVC
  • Theorem 1
  • Lemma 2: Properties of NF lavalle2006planning
  • Lemma 3: Convergence pierson2017distributed