Combining evolutionary algorithms and AI to design medical actuators

Abstract: The characteristics of soft robots make them better candidates for applications such as healthcare against their traditional counterparts, due to their enhanced safety, adaptability and trivial human-robot interaction. Different actuating systems have been proposed for soft robotics, with the most interesting one being biohybrid actuators. On the other hand, since this technology is young, the designing process of biohybrid actuators is not well formalized. To enhance the applicability of this type of actuators, the utilization of a NeuroEvolution in the design process is studied. Candidate solutions are encoded by artificial neural networks and evolved under NeuroEvolution terms. These automatically produced biohybrid sketches are tested within a simulated environment to identify the suitability of this methodology. Whereas the best candidates can be delivered in the following steps of the designing process for further fine tunning based on the specific application.