Human-Robot Interaction Control Based on Motion Intent Estimation

Abstract This paper presents an adaptive impedance control framework for human-robot cooperation under dual uncertainties of unknown human motion intent and uncertain robot dynamics. Human motion intention is denoted by the desired trajectory within the kinematic model of the human limb. Obtaining this trajectory poses notable challenges due to the highly nonlinear and time-varying dynamics inherent in human limb modeling. To address this issue, a novel online estimation architecture is developed by integrating Radial Basis Function Neural Networks (RBFNN). This adaptive impedance controller integrates estimated motion intent to enable real-time tracking of a predefined impedance reference model. This developed framework’s active collaboration with its human partner can be demonstrated by experimental results. This work provides novel theoretical insights and practical contributions to the field of human-robot interaction.