Path Planning for Deep-Sea Nodule Collection: Obstacle and Path-Overlap Avoidance

Abstract

Deep-sea polymetallic nodule collection requires computationally efficient path planning for a tethered collector operating in large and obstacle-rich seafloor environments. The collector must avoid static obstacles while accounting for nonholonomic motion constraints, tether-related limits, and the need to minimise repeated traversal of already harvested terrain. The existing Allseas Motion Primitive 𝐴* planner can generate feasible routes, but it becomes computationally expensive on large maps and produces overlapping obstacle-avoidance paths. Repeated traversal is undesirable because it yields no additional resource recovery and may degrade the local soil structure, increasing the risk of reduced collector mobility.

This thesis develops a harvest-history-aware path-planning framework, 𝐴𝑙𝑙*, for deep-sea nodule collection. The proposed method uses a cost-weighted grid-based 𝐴∗ planner that incorporates obstacle costs, a vessel-line guidance, heading-deviation restrictions, and an overlap penalty applied to previously harvested terrain. A Gaussian overlap-penalty formulation is used to discourage repeated traversal while retaining flexibility near the edges of previous tracks. The framework further includes reachability analysis, an expanded neighbourhood structure, and a motion-control layer to generate smooth trajectories and guide the collector towards the required terminal heading.

The proposed framework was evaluated against the Allseas Motion Primitive 𝐴∗ baseline in three static simulation scenarios with different map sizes and obstacle configurations. Performance was assessed using total pickup, total duration, field efficiency, overlap severity, and average planner execution time. Across the three tested scenarios, the proposed method increased total pickup by 2.98–6.23% and field efficiency by 2.63–2.86%. The overlap severity index was reduced by 18.99–26.26%, indicating that obstacle-avoidance paths were distributed more evenly across the available terrain. Average planner execution time was reduced by 94.71–96.59%, while total route duration increased by only 0.79–3.27%.

Overall, incorporating harvest history directly into the planning cost map improved harvesting performance, reduced overlap, and substantially reduced planning time in the tested simulation environments. The framework provides a basis for further development of path-planning methods for tethered deep-sea collection vehicles. Future work should investigate adaptive parameter tuning, alternative collection strategies, higher-fidelity tether and geotechnical constraints, and online replanning under uncertain terrain information.