A Stiffness-Fault-Tolerant Control Strategy for Reliable Physical Human-Robot Interaction.

Elastic actuators allow to specify the characteristics of physical human-robot interactions and increase the intrinsic safety for the human. To ensure the reliability of the interaction, this paper investigates detection and compensation of stiffness faults. A recursive least squares algorithms is used to detect faults and obtain an estimation of the actual stiffness value online. An adaptation law based on the estimation is proposed to adjust parameters of an impedance control to maintain a desired interaction stiffness. A simulation of an exemplary elastic actuator shows that the developed stiffness-fault-tolerant control strategy achieves a dependable human-robot interaction.