Multi-USV Coverage Path Planning Using Spatial Graph Multi-Actor-Attention-Critic Reinforcement Learning Framework With Operator Pooling