Comparing model-based control methods for simultaneous stiffness and position control of inflatable soft robots

Abstract

Inflatable robots are naturally lightweight and compliant, which may make them well suited for operating in unstructured environments or in close proximity to people. The inflatable joints used in this article consist of a strong fabric exterior that constrains two opposing compliant air bladders that generate torque (unlike McKibben actuators where pressure changes cause translation). This antagonistic structure allows the simultaneous control of position and stiffness. However, dynamic models of soft robots that allow variable stiffness control have not been well developed. In this work, a model that includes stiffness as a state variable is developed and validated. Using the stiffness model, a sliding mode controller and model predictive controller are developed to control stiffness and position simultaneously. For sliding mode control (SMC), the joint stiffness was controlled to within 0.07 Nm/rad of a 45 Nm/rad command. For model predictive control (MPC) the joint stiffness was controlled to within 0.045 Nm/rad of the same stiffness command. Both SMC and MPC were able to control to within 0.5° of a desired position at steady state. Stiffness control was extended to a multiple-degree-of-freedom soft robot using MPC. Controlling stiffness of a 4-DOF arm reduced the end-effector deflection by approximately 50% (from 17.9 to 12.2cm) with a 4 lb (1.8 kg) step input applied at the end effector when higher joint stiffness (40 Nm/rad) was used compared with low stiffness (30 Nm/rad). This work shows that the derived stiffness model can enable effective position and stiffness control.