REACT: Real-time Entanglement-Aware Coverage Path Planning for Tethered Underwater Vehicles

TL;DR

This work addresses tether entanglement in underwater inspection by introducing REACT, a real-time entanglement-aware coverage path planning framework. It offline-constructs an SDF map from point clouds and computes nominal waypoints, then online enforces a tether-length constraint $L_{ ext{max}}$ to prevent entanglement and provides entanglement-free waypoint updates. Key contributions include a taut-tether model with efficient node-based shortcutting, a two-mode entanglement-aware planner with a disentanglement search and a recovery-path refinement, and an MPC-based controller integrating with the offline planner. Realistic simulations and real-world experiments show that REACT maintains safety and complete coverage while reducing post-mission detangling, outperforming a baseline CPP that lacks tether awareness.

Abstract

Inspection of underwater structures with tethered underwater vehicles is often hindered by the risk of tether entanglement. We propose REACT (real-time entanglement-aware coverage path planning for tethered underwater vehicles), a framework designed to overcome this limitation. REACT comprises a computationally efficient geometry-based tether model using the signed distance field (SDF) map for accurate, real-time simulation of taut tether configurations around arbitrary structures in 3D. This model enables an efficient online replanning strategy by enforcing a maximum tether length constraint, thereby actively preventing entanglement. By integrating REACT into a coverage path planning framework, we achieve safe and entanglement-free inspection paths, previously challenging due to tether constraints. The complete REACT framework's efficacy is validated in a pipe inspection scenario, demonstrating safe navigation and full coverage inspection. Simulation results show that REACT achieves complete coverage while maintaining tether constraints and completing the total mission 20% faster than conventional planners, despite a longer inspection time due to proactive avoidance of entanglement that eliminates extensive post-mission disentanglement. Real-world experiments confirm these benefits, where REACT completes the full mission, while the baseline planner fails due to physical tether entanglement.

Figures (7)

• Figure 1: Overview of REACT: Offline, an SDF map is generated from a point cloud, and an off-the-shelf CPP feng2024fc is used to compute a near distance-optimal waypoint sequence. Online, REACT provides an entanglement-free next waypoint to the controller.
• Figure 2: Tether shortcutting during ROV motion from $t_1$ to $t_2$. (1) Initial tether with new ROV position $\mathbf{p}_{\text{rov}}$ appended; (2) Successful shortcut from the end node; (3) Collision encountered when attempting further shortcutting, skipping to the next node; (4) Another collision detected from the new node; (5) Successful shortcut from a subsequent node; (6) Final tether configuration (yellow) after applying all feasible shortcuts.
• Figure 3: Disentanglement path search process. At time $t$, the tether configuration is $\mathbf{P}_{\text{tether}}(t) = \{\mathbf{p}_0, \mathbf{p}_1, \mathbf{p}_2, \mathbf{p}_3\}$. (1) The shortest path $\mathbf{P}_{s3}$ from $\mathbf{p}_3$ to the waypoint $\mathbf{W}[k]$ is appended to the tether path, forming the augmented path $\mathbf{P}_{\text{tether}}^{(0:3)} \cup \mathbf{P}_{s3}$, which is passed to the tether model to compute $\mathbf{P}_{\text{tether}}(t+1)$. The resulting tether length exceeds $L_{\max}$, indicating a violation. (2) The pivot shifts to $\mathbf{p}_2$, and the augmented path $\mathbf{P}_{\text{tether}}^{(0:2)} \cup \mathbf{P}_{s2}$ is evaluated, but still violates the constraint. (3) With pivot at $\mathbf{p}_1$, the augmented path $\mathbf{P}_{\text{tether}}^{(0:1)} \cup \mathbf{P}_{s1}$ yields a feasible configuration $\mathbf{P}_{\text{tether}}(t+1)$ with $L_{\text{tether}} \leq L_{\max}$. (4) The recovery path $\mathbf{P}_{\text{recovery}}$ is thus executed, consisting of $\mathbf{P}_{\text{r}1}$, tracing back from $\mathbf{p}_3$ to $\mathbf{p}_1$, and $\mathbf{P}_{\text{r}2} = \mathbf{P}_{s1}$, leading to the waypoint $\mathbf{W}[k]$.
• Figure 4: Comparison of coverage and tether length.
• Figure 5: 3D views of final trajectories: (a) CPP results in entangled tether path, (b) REACT yields a non-entangled tether path, reflecting effective entanglement avoidance.

Theorems & Definitions (2)

• Remark 1
• Remark 2