Simulation Study of Current-Aware Adaptive Potential-Field Local Planning for a Planar ROV Model in Time-Varying Currents

Abstract

This simulation study evaluates a current-aware local planner for a circular planar remotely operated vehicle model in known static maps with time-varying currents. PFPA-v2 combines bounded attraction, threat-weighted repulsion, current-adjusted local commands, and prevalidated connected-cluster escape. A common continuous audit applies the same obstacle-body and inset-boundary clearance criterion to planned and executed trajectories, while a planar 3-DOF LOS–PI layer separates reference-path behavior from closed-loop tracking. In a frozen 24 planning factorial over 60 map–current scenarios, PFPA-v2-primary yielded 54/60 safe references versus 44/60 with the four tested components disabled. The paired integration-level difference was significant, whereas no individual main or two-factor effect survived multiplicity correction. In an independently tuned common-clearance comparison, CA-Informed RRT* showed higher observed reliability and shorter conditional paths, whereas PFPA-v2-P09 produced greater obstacle clearance and much lower recorded desktop planning time; map-cluster intervals did not support a planner-family ranking. Current-aware execution was more reliable than no-current-feedforward execution in the unified PFPA-v2-primary chain. Moderate-current and affine-remapped observational-current studies were exploratory. The results support an interpretable simulation benchmark, not global completeness, six-degree-of-freedom feasibility, or field-scale ROV validation.