Evolutionary Optimization of Soft Swimming Robots

Engineering a soft robotic swimmer presents a unique challenge due to the inherently large numbers of degrees of freedom associated with their design. We present a method of developing optimal designs for soft swimmers powered by artificial muscle. To do so, the U-NSGA-III multi-objective optimization method is coupled with a FD/DLM FSI method simulating robotic fish actuated by an artificial muscle model. Since U-NSGA-III is a multi-objective method it in effect produces a map of potentially optimal solutions. This information can then be employed in the design of robotic swimmers as well as the research of swimming methods. We successfully evolve multiple, unique swimmers that are biomimetic in nature from an initial set of randomly selected design parameters, in essence reproducing a small piece of natural evolution.