A Computationally-Efficient Rollout-Based Approach for Bathymetric Mapping with Multiple Low-Cost Unmanned Surface Vehicles

This paper proposes an integrated software-hardware approach for automated bathymetry mapping using low-cost unmanned surface vehicles (USVs). The solution takes inspiration from adaptive sampling, aiming to minimize the time, hardware, and training costs compared to traditional coverage path planning (CPP) methods. The proposed approach implements a scalable Markov Decision Process (MDP)-based model and a planning method based on multi-agent approximate dynamic programming (ADP), which prescribes USVs to seek the most informative samples in bandwidth-limited environments with uncertain and incomplete prior map information. The control structure offloads surrogate model updates and path planning to the control station and ensures mission completion even with communication dropouts. Computer simulation results on both real-world and computer-generated bathymetry data demonstrate the effectiveness of the proposed approach in terms of predictive accuracy and efficiency. The proposed solution has the potential to significantly aid coastal development by streamlining access to accurate models of near-shore bathymetry.