Monitoring underwater ecosystems is critical for assessing marine biodiversity and tracking the ongoing impacts of global climate change. Bio-inspired soft robots present an ideal solution for these environments, leveraging morphological mimicry, physical compliance, and low-acoustic locomotion to enable the non-disruptive observation of aquatic life. In this thesis, we investigate a squid-inspired, jet-propelled robotic platform, presenting contributions in its mechatronic design, performance analysis, and autonomous motion control.

Specifically, we developed a variable-volume, jet-propelled robot. The robot simulates the pulsatile dynamics of biological squids; its consistent kinematics and modular design enable it as a versatile testbed for studying complex fluid-structure interactions and validating motion control frameworks. By leveraging the system’s modularity, we adapt the robot to emulate the jetting mechanics of a biological salp and conducted an experimental analysis on the influence of unidirectional versus bidirectional flow regimes on the platform’s propulsive efficiency. Based on the experiment results, we formulated design guidelines to decide the selection between unidirectional and bidirectional flow regimes in jet-propelled aquatic systems. Finally, to achieve autonomous 2D planar locomotion, we integrated a steerable nozzle into the testbed architecture. Considering the highly nonlinear dynamics of the platform, we developed a Reinforcement Learning (RL) motion controller. This control policy was initially trained and validated within a custom dynamic simulation environment before being successfully deployed onto the physical hardware, demonstrating the efficacy of the closed-loop control framework.