Purpose
To ensure accurate position and force control of massage robot working on human body with unknown skin characteristics, this study aims to propose a novel intelligent impedance control system.
Design/methodology/approach
First, a skin dynamic model (SDM) is introduced to describe force-deformation on the human body as feed-forward for force control. Then a particle swarm optimization (PSO) method combined with graph-based knowledge transfer learning (GKT) is studied, which will effectively identify personalized skin parameters. Finally, a self-tuning impedance control strategy is designed to accommodate uncertainty of skin dynamics, system delay and signal noise exist in practical applications.
Findings
Compared with traditional least square method, genetic algorithm and other kinds of PSO methods, combination of PSO and GKT is validated using experimental data to improve the accuracy and convergence of identification results. The force control is effective, although there are contour errors, control delay and noise problems when the robot does massage on human body.
Originality/value
Integrating GKT into PSO identification algorithm, and designing an adaptive impedance control algorithm. As a result, the robot can understand textural and biological attributes of its surroundings and adapt its planning activities to carry out a stable and accurate force tracking control during dynamic contacts between a robot and a human.