Abstract: Traditional communication systems and conventional channel estimation methods are often inadequate for effectively mitigating the severe Doppler frequency shifts caused by the rapid movement of underwater platforms in acoustic communication scenarios. Starting from the characteristics of underwater acoustic channels in mobile environments, this paper investigates OTFS (orthogonal time frequency space) modulation for underwater acoustic communication through theoretical analysis and simulation studies. A simulated underwater acoustic channel and communication framework is developed to compare the performance of OFDM (orthogonal frequency division multiplexing) and OTFS in dynamic underwater conditions, thereby demonstrating the necessity and feasibility of deploying OTFS in communication systems for high-speed AUVs (autonomous underwater vehicles). To further enhance the system performance, a channel estimation method based on variational inference and sparse Bayesian learning is evaluated against conventional approaches, highlighting its advantages within the OTFS framework. Moreover, the proposed OVSB (optimized variational sparse Bayesian) estimation method is integrated into a practical OTFS-based underwater acoustic communication system, and experimental results confirm the accuracy of simulation predictions.