This chapter addresses the integrated design of communication and control for maritime unmanned systems (MUSs) in challenging underwater environments. Two key tasks are investigated: (1) communication-efficient and collision-free motion planning for single underwater vehicles in fading channels, and 2) optimally persistent formation control for multiple AUVs with unknown interaction topology. For the first task, a model-based IRL estimator is developed to predict the stochastic signal-to-noise ratio (SNR) in fading channels. Based on the predicted SNR, an integrated optimization framework is constructed by considering vehicle dynamics, communication capacity, collision avoidance, and position control. A model-free IRL algorithm is then designed to solve this optimization problem. For the second task, a local topology estimator is first designed to infer the unknown interaction relationships among AUVs. An optimally persistent topology is then generated using graph theory to balance energy efficiency and topology connectivity. Finally, a model-free inverse reinforcement learning formation controller is developed to maintain the desired formation shape. The effectiveness of the proposed approaches is validated through both simulations and experiments, demonstrating their superior performance compared to existing solutions.