NMPC-based variable admittance control with passivity constraints

When robots interact with their environment, managing the physical interaction safely and accurately is critical. Impedance and admittance control allow such interaction and have been fundamental to robotic control for decades. For challenging applications, it is often necessary to implement a time-varying impedance profile, which necessitates more complex conditions to guarantee stability. This paper introduces a method to implement variable admittance control with explicit passivity constraints using non-linear model predictive control (NMPC). The dynamics of the impedance parameters are directly included in the predictive controller. Contrary to previous approaches, the computed control signals enforce simultaneously the passivity condition and the constraints, on both the robot state and the control signals.