提出可微分弹性机器鱼模型，实现快速仿真与高效设计优化。

Body flexibility plays a critical role in fish-like swimming, as the spatial distribution of stiffness governs body deformation, hydrodynamic loading, and propulsive performance. Exploiting this mechanism in robotic fish requires dynamic models that capture continuous body elasticity, fluid-structure interaction, and the resulting self-propelled motion. Existing approaches often prescribe body kinematics, approximate the body using discrete rigid or compliant segments, or incur high computational costs that limit their use in design optimization. In this letter, we present a differentiable full-body dynamics model and fast simulation framework for motor-actuated elastic robotic fish based on Hamilton's principle. The proposed formulation represents the robot as a continuously deformable elastic body and couples its structural dynamics with hydrodynamic forces without prescribing body kinematics. The resulting simulator is differentiable with respect to model and design parameters, enabling efficient gradient-based optimization. Numerical convergence studies and experiments with a physical robotic fish validate the proposed framework. Finally, gradient-based optimization of the body stiffness distribution demonstrates its utility for efficient design of elastic robotic fish.