Multi-UAV assisted cross-boundary communication scheme for AUV swarms via multi-agent reinforcement learning approach

Abstract

In maritime emergency response operations, autonomous underwater vehicles (AUVs) are frequently deployed for underwater search and marine data collection missions. However, establishing real-time water-air cross-boundary communication with AUV remains a crucial challenge. Traditional deployment of surface methods faces limitations, such as unreliable and imbalanced connections, especially when AUVs are tasked with covering large, dynamic search areas. To address the challenge, this paper proposes a novel multiple unmanned aerial vehicles (UAVs) collaboration communication scheme, in which the UAVs utilize hydrophones as mobile base stations to establish water-air cross-boundary communication with AUV swarms. First, we develop a communication coverage and energy consumption model for UAVs. Then, we introduce an AUV position prediction algorithm based on particle filter (PF), which estimates the position state information of AUVs in real time while reducing the frequency of dynamic adjustment by UAVs. Finally, we formulate the dynamic deployment of UAVs as a partially observable Markov decision process (POMDP) to optimize communication performance and energy consumption, and propose a dynamic deployment scheme based on multi-agent deep deterministic policy gradient (MADDPG) to deal with the coverage imbalance problem and provide maximum coverage service. Extensive simulations demonstrate that the proposed scheme can reduce the energy consumption by about 7.9% compared to the no-prediction scheme, effectively balancing coverage fairness and energy consumption while satisfying the communication requirements of AUV swarms.