Jellyfish-inspired Robots Enabled by Soft and Hard Actuators

Dielectric elastomer (DE) materials, a category of electroactive polymers, can convert an electrical input into mechanical work. They can be used to design Dielectric elastomer actuators (DEAs) that are flexible, resilient, lightweight, and durable and provide such properties without suffering high financial costs, which makes them particularly promising for soft robotic applications. Unfortunately, severe technical deficiencies limit the development of DEAs: hard-to-embed algorithms for self-sensing, low output forces, and challenging integration with other actuators. This thesis provides some approaches to address these limitations by combining various DEAs applications. First, a tubular DEA is designed to develop a self-sensing algorithm by building a model between the capacitance and measured displacement. Fast Fourier Transform (FFT) is used to filter a given frequency of the probing current and voltage, and then calculate the capacitance at the probing frequency with the probing current and voltage during each time window. With the relationship between displacement and the capacitance of the DE tube, the movement of the DE actuator can be estimated online and achieve self-sensing without an external sensor. Second, a DEA-enabled robotic jellyfish robot is developed based on contracting the muscle-like behavior of DE material. It combines a DE diaphragm actuator with a transmission mechanism, which can provide a compliant thrust force to propel the jellyfish robot to transit through the water. A data-driven model is developed to capture the vibration in the first step. The process of contracting the bell and producing thrust force is captured by a physical model in the second step. Lastly, a novel 2D maneuverable jellyfish robot fabricated with multiple DE membranes and IPMC is introduced, which uses the DE membrane to generate a periodic contraction on its eight fins to provide propulsion. The robot utilizes an IPMC to generate a bending moment that directs the heading angle of its swimming. Lastly, a biomimetic jellyfish robot driven by the DC motor is fabricated and activated to mimic real locomotive behavior, which has a higher speed and better controllability compared with the previous two robotic jellyfish. Due to the nonlinear fluid dynamics and electromechanical coupling, a model-free control method based on reinforcement learning (RL) is employed to offer powerful algorithms to search for optimal controllers of systems.