Actuation-aware configuration manifold generation for path planning of cable-driven snake-like robots in confined utility tunnel

Abstract Urban utility tunnels pose severe navigational challenges for cable-driven snake-like robots (CDSRs) due to narrow corridors, sharp bends, and obstacles. Because of strict physical limitations, successful head navigation does not guarantee acceptable whole-arm traversal. To address this, an actuation-aware configuration manifold generation paradigm is proposed for CDSR path planning. A two-stage framework evaluates the discrete backbone: envelope nonintrusion, curvature limits, and tension-risk bounds define hard feasibility, while safety clearance and curvature gradients provide soft ranking for feasible candidates. At the planning level, a harmonic potential field provides a global descent direction devoid of spurious interior minima, while a centerline streamline prior offers topological branch guidance within the orthogonal complement of the potential-field gradient. Subsequently, a follow-the-leader mechanism translates head guidance into a planning-level acceptable whole-arm configuration evolution. Extensive simulations across six simple channel types and a complex utility-tunnel topology demonstrate improved performance of the proposed framework under the evaluated simulation conditions. In the complex topology, the full method achieves both a task success rate and an acceptable success rate of 86%, compared with artificial potential field and harmonic field methods, which suffer 0% success rates due to collision or stagnation. Ablation studies demonstrate that omitting curvature or tension constraints causes severe overbending (exceeding 80°) or elevated tension risks. Ultimately, the proposed method shifts the planning focus from geometric head-arrival trajectories to the generation of safe, planning-level acceptable whole-arm configuration sequences in highly constrained environments.