Model Predictive Control of a Soft Laparoscope Using Neural Networks

Soft robots are an emerging alternative to traditional surgical equipment. Specifically, capabilities of these like flexibility and controlled rigidity, are exploited in enclosed environments where requirements can restrict the use of rigid apparatus. In this work, a soft robot intended to perform laparoscopic activities is used. The design of a cascade control system for this soft laparoscope is presented, composed of two control algorithms: one in charge of controlling the position of the robot and the second in charge of controlling the pressure on its inner chambers. The latter is based on a standard ON-OFF controller, and the former is based on the Model Predictive Control (MPC) methodology, where Deep Neural Networks (DNN) are used for modeling the actuator. The results show that our control system allows the soft laparoscope to reach a desired orientation with a maximum error of 0.06 rad.