Pattern-Aware Task Assignment and Depth-Based Deconfliction for Multi-AUV Oceanic Search Operations

The deployment of multiple Autonomous Underwater Vehicles (AUVs) for large-scale oceanic search and mapping missions is frequently constrained by inefficient task assignment and strict collision avoidance protocols. Traditional deconfliction methods block adjacent operational regions, significantly reducing the feasible search area and delaying the exploration of high-priority zones. Although this limitation is widely acknowledged in recent surveys, the direct integration of physical search-pattern geometry into the assignment constraints themselves remains largely unexplored. To address this gap, a mathematical model for dynamic task assignment executed across macro- and micro-cycles is proposed. The objective is formulated to maximize search rewards based on finding likelihood and temporal efficiency. To bypass the restrictive nature of neighboring-region exclusion constraints, a novel integration of alternating narrowing and widening spiral patterns is introduced, combined with depth-based layer assignment for transiting AUVs. The relaxation is validated at two levels. At the constraint level, systematic Monte Carlo simulations over 50 trials show that the pattern-aware method attains a statistically significant reward improvement over a standard strict-adjacency baseline (paired t(49)=7.15, p<0.001) while remaining free of assignment-level conflicts. At the trajectory level, the narrowing and widening sweeps and the depth-layered transits are explicitly simulated and scored with a single physical conflict metric applied identically to every method. This analysis shows that the relaxed constraint alone is not physically sufficient: executing the alternating spiral geometry reduces sonar-interference events from 2.36 to 0.52 per run, and an additional phase-admissibility condition is required to eliminate them. A regime analysis establishes that the mechanism is valid when the region side exceeds approximately four sonar ranges. The results indicate that physical search-pattern geometry can be embedded in assignment constraints to recover most of an unconstrained heuristic’s early-discovery advantage while remaining deconflicted, within an explicitly characterized operating envelope.