Real-time adaptive impedance compensator using simultaneous perturbation stochastic approximation for enhanced physical human–robot interaction transparency

In the physical human–robot interaction (pHRi) system, the human and the robot are physically coupled, and it makes the human and the robot always influence each other. In engineering tasks, using a force sensor is the norm for control the robot through the interaction force between the human and the robot. However, the force measured from the force sensor contains the force intended by human motion and the natural force feedback generated by the robot movement and the human hand impedance due to the coupled dynamics. Therefore, it is necessary to characterize the human hand dynamics to improve transparency. However, it is difficult to estimate the human hand impedance, which is the primary source of natural force feedback in real-time. This paper proposes a real-time adaptive hand impedance compensator to enhance transparency in various pHRi conditions with human hand dynamics. The proposed algorithm regulates the impedance compensator’s parameters to find optimal values that minimize the energy-based cost function using Simultaneous Perturbation Stochastic Approximation (SPSA) and AMSGrad. SPSA is a useful method when the exact relationship between the parameters and the cost function is unknown. AMSGrad is a state-of-the-art technique widely used as an adaptive learning rate method in deep learning fields. The proposed real-time adaptive impedance compensator decreases the influence of natural force feedback by updating the parameter appropriately depending on the pHRi conditions, thus improving the transparency.