Coverage Control for Underwater Acoustic Mobile Sensor Networks with Connectivity and Energy Constraints Based on Multi-Agent Deep Reinforcement Learning

Abstract

Three-dimensional coverage deployment is a fundamental challenge for underwater mobile sensor networks (UMSNs), where the coupled requirements of high coverage, guaranteed connectivity, and low energy consumption must be simultaneously satisfied under dynamic and partially observable conditions. To address these issues, this paper proposes a coverage-prioritized constrained multi-agent proximal policy optimization (CP-CMAPPO) algorithm that decomposes the control process into three hierarchical layers: a coverage-frontier target allocation layer for explicit exploration guidance, a residual MAPPO control layer for local motion refinement, and a connectivity-aware action synthesis layer for topology integrity maintenance. The simulations demonstrate that CP-CMAPPO achieves nearly complete cumulative coverage while guaranteeing full network connectivity. Notably, the proposed method maintains stable and predictable energy expenditure despite the additional cost of ensuring coverage and connectivity. These results validate that the proposed hierarchical architecture effectively mitigates the exploration burden, prevents topology fragmentation, and enables robust, energy-efficient cooperative coverage in underwater acoustic mobile sensor networks.