This chapter presents an integrated framework for the co-design of detection, communication, and control in intelligent underwater systems, aiming to bridge the gap between autonomous perception and human-in-the-loop decision making. The proposed integration enables underwater vehicles to jointly sense, transmit, and act under harsh and bandwidth-limited acoustic environments, improving spectrum efficiency, energy utilization, and operational robustness. The chapter first establishes an integrated design philosophy that embeds signal integration and learning-based control within a cyber-physical architecture. By employing orthogonal chirp division multiplexing (OCDM) waveforms, the system achieves simultaneous target detection and self-localization, effectively merging sensing and communication functions. On top of this, an RL-based control strategy coordinates motion planning, obstacle avoidance, and task execution, ensuring adaptive and collision-free navigation. Furthermore, to extend autonomy toward complex manipulation tasks, a human-on-the-loop teleoperation paradigm is introduced, integrating acoustic-radio cross-medium communication, vision-assisted localization, and IRL-based control for transparent bilateral interaction between operator and AUV. Field experiments in both pool and marine environments validate the effectiveness of this integrated framework.