Line-of-Sight-Constrained Multi-Robot Mapless Navigation via Polygonal Visible Regions

Abstract

Multi-robot systems rely on underlying connectivity to ensure reliable communication and timely coordination. This paper studies the line-of-sight (LoS) connectivity maintenance problem in multi-robot navigation with unknown obstacles. Prior works typically assume known environment maps to formulate LoS constraints between robots, which hinders their practical deployment. To overcome this limitation, we propose an inherently distributed approach where each robot only constructs an egocentric visible region based on its real-time LiDAR scans, instead of endeavoring to build a global map online. The individual visible regions are shared through distributed communication to establish inter-robot LoS constraints, which are then incorporated into a multi-robot navigation framework to ensure LoS-connectivity. Moreover, we enhance the robustness of connectivity maintenance by proposing a more accurate LoS-distance metric, which further enables flexible topology optimization that eliminates redundant and effort-demanding connections. The proposed framework is evaluated through extensive multi-robot navigation and exploration tasks in both simulation and real-world experiments. Results show that it reliably maintains LoS-connectivity between robots in challenging environments cluttered with obstacles, even under large visible ranges and fragile minimal topologies, where existing methods consistently fail. Ablation studies also reveal that topology optimization boosts navigation efficiency by around $20\%$, demonstrating the framework's potential for efficient navigation under connectivity constraints.

Figures (9)

• Figure 1: (a) Four robots navigate in a simulation environment cluttered with small irregular obstacles while always maintaining LoS-connectivity ($100m\times 50m$). (b) The trajectories of four robots and snapshots of their connectivity graphs when exploring an initially unknown garage environment 10577228 ($87m\times 69m$). (c) Three robots exploring an outdoor forest environment adapted from cao2023representation ($86m\times 95m$).
• Figure 2: Framework of LoS-connectivity constrained multi-robot navigation method. Each robot independently constructs its visible region with the polygonal approximation, based on real-time LiDAR scans. After sharing with immediate neighbors, three types of connectivity constraints are formulated as the weight of the robots' connectivity graph, which is then proceeded with topology optimization. Next, the masked graph Laplacian matrix is used to derive the connectivity velocity $\boldsymbol{u}_{i}^{c}$, which is then fused with external navigational velocity $\boldsymbol{u}_{i}^{n}$ through the role-based scaling module, following the design in bai2025realmrealtimelineofsightmaintenance. Finally, each robot is driven independently by the fused velocity command to maintain LoS-connectivity between robots while navigating to its own target. The navigational velocity is obtained from a mapless path planner proposed in yang_FARPlanner_2022. The framework can also be extended for multi-robot exploration by integrating a multi-robot mapper for frontier extraction and a navigation-target assigner, as exemplified in Fig. \ref{['fig_wide_grid']}(b)(c).
• Figure 3: (a): A robot (denoted by $+$) with its visible region and polygonal approximation, whose boundaries are highlighted by purple and green dotted lines in (b) and (c), respectively. The radial angle-based interpolation adapts to the boundary geometry by placing denser points (green dots in (a)) along boundaries of high curvature and sparser points along flatter boundaries. (b) and (c) also show the gradient field of $d^{\text{hull}}$ (purple arrows) and $\tilde{d}^{\text{los}}$ (green arrows) respectively, in a zoomed-in area. A probe point (denoted by $\blacktriangle$) in (a) has the exact $d^{\text{los}} = 1.08m$ and $\tilde{d}^{\text{los}} = 1.07m$, while $d^{\text{hull}} = 12.50m$ showing significant metric error. Moreover, the gradient of $d^{\text{hull}}$ tends to trap a robot within a small radial sector.
• Figure 4: (a) Error distribution of different LoS-distance metrics at uniformly sampled points in a visible region shown in Fig. \ref{['fig_linear_approx']}, under various flipping radius $r_{\text{flip}}$. Here we set $\Delta\theta = 0.5^{\circ}$. (b) and (c): Ablation study of using Topo-Opt when four robots are exploring a garage environment in Fig. \ref{['fig_wide_grid']}(b), where (b) compares the temporal variation of $\lambda_2$ and (c) shows the histogram of the number of connections between four robots during exploration.
• Figure 5: Relative time and distance efficiency of compared methods with four robots navigating in a cluttered environment shown in Fig. \ref{['fig_wide_grid']}(a). Dashed lines indicate the average performance. The indices of different runs are sorted in ascending order based on the total distance to the targets. (a) and (b): only one robot is assigned a random target; (c) and (d): all four robots have their targets.

Theorems & Definitions (12)

• Definition 1: Visible Region
• Proposition 1: Visibility Determination katz2007direct
• Definition 2: Line-of-Sight-Distance
• Proposition 2
• Proposition 3
• Proposition 4
• Remark 1
• Remark 2
• proof
• Corollary 1