Soft robotics may be utilized to create new, more lifelike types of locomotion as well as to comprehend complex biomechanics using robotic model platforms. We built a fully soft robotic subcarangiform fish based on prior results on undulatory movement in fish, using antagonistic fast-PneuNet actuators and hyperelastic eutectic Gallium-Indium (eGaIn) implanted in silicone channels for strain sensing. These sensors are capable of measuring curvature because the electrical conductivity of the liquid metal reduces according to the silicone structure's stretching as the robot bends. As with its natural model, it advances in a wave-like pattern, with all components having a stiffness comparable to that of a fish body. 
To establish control, a simple, data-driven lumped parameter approach is proposed. This approach enables accurate yet lightweight computations that are customized using experimental data and a genetic algorithm. The model accurately predicts the robot's behavior over a wide range of driving frequencies and pressure amplitudes, taking into account the influence of the soft actuators' antagonistic co-contraction. The model is used to prototype an amplitude controller that is then deployed to the robot in order to reach the setpoint of a tail-beat amplitude using fully soft and real-time strain sensing. Knowing when a fish bends its caudal fin during swimming enables researchers to develop even more complex and life-like swimming robots.