Control of a pneumatically actuated, fully inflatable, fabric-based, humanoid robot

Although humanoid robots take the form of humans, these robots often approach manipulating the world in a very different way than humans. For example, many humanoid robots require precise position control and geometric models to interact successfully with the world. Humanoid robots also often avoid making contact with the world unless the contact can be well modeled. In this work, we present preliminary results on soft robot platforms that can change the way humanoid robots interact with humans and human environments. We present preliminary control methods and testing on fully inflatable, pneumatically actuated, soft robots. We first show that model predictive control (MPC) and linear quadratic regulation (LQR) are sufficient for position control of a single joint with one degree of freedom. We also demonstrate MPC and LQR as methods of control for an inflatable humanoid robot on one arm using five degrees of freedom. Our initial development for multi-joint control is based on the methods developed for the single degree of freedom platform. Using the MPC controller with joint space commands, a task of picking up a board from a chair and placing it in a box was successful eight out of ten times. Our models and control methods will allow for a new type of humanoid robots that are well suited to interacting more safely and naturally in human environments.