Force Control Beyond the Passivity Interaction Paradigm

Several interaction control algorithms have been proposed for physical human-robot interaction (pHRI) with the aim of assisting human movements. Some example are “path control” [4], “virtual model control” [7], “clime” [9] or gravity compensation [5], [3]. This class of algorithms is often built upon a force/torque controller that is supposed to be ideally fast. However when the algorithm is implemented the force variable cannot be commanded instantaneously and its transients depend both on robot and human dynamics. In particular, using a traditional controller we can have slow force/torque responses when the human displays a low mechanical impedance and force/torque overshoots when the human has a high impedance configuration. These effects are undesirable because they can generate disturbances that interfere with the rehabilitation strategy and even potentially lead to instability and injuries.