Game theory based vision impedance control for human-robot interaction

To address the influence of multiple forces on the accuracy of position tracking in human-robot interaction systems, this paper proposes a vision impedance control method based on game theory. With the end-effector position obtained by vision feedback and the adaptive impedance law, the human-robot interaction force can be estimated. The position error is then converted into a constraint force to ensure the output position remains within a specified limit. The preset control force, human-robot interaction force, and constraint force are regarded as participants and a multi-party cooperative differential game algorithm is developed to derive the optimal impedance controller for the end-effector. The convergence of the position error is proved using Lyapunov functions. The performance of proposed method is validated through simulations and experiments. The proposed method can flexibly adjust the proportions of various complex forces on the end-effector during the human-robot interaction. During the interaction phase under multiple complex forces, the mean squared error of the end-effector position with the proposed method is merely 7.7 % and 6.0 % of those obtained with the sensorless force estimation-based control and the repetitive impedance learning-based control, respectively. Meanwhile, it can reduce the computation complexity of conventional vision methods and improve the tracking accuracy within the constraints.