A non contact, wide-ranging approach to human–robot interaction based on broad learning system

Purpose Haptic enables humans to experience the scenes of touch, which is important for human–robot interaction (HRI). However, most existing haptic devices implement such scenes of touch in a direct contact manner, the friction and inertia present during the interaction reduces the naturalness, hindering the flexibility of the operation. This paper aims to provide a noncontact, wide-ranging force feedback approach to HRI possessing enhanced naturalness and immersion. Design/methodology/approach In this paper, the authors use an array of electromagnetic coils to generate a multidimensional electromagnetic force. Because this electromagnetic force can hardly be computed efficiently and accurately by electromagnetism, a computationally efficient model is introduced to solve this problem. The paper proposes a novel innovative noncontact natural human–robot interface based on elastic network broad learning system with gate recurrent unit. In addition, the paper introduces the Google MediaPipe Hands for markerless adaptive gesture tracking to overcome the problem of narrow interaction space in conventional gesture-based interaction methods. Findings The experimental results show that the proposed method is able to fit the target force to achieve noncontact real-time force feedback. According to the evaluations of the volunteers, the provided forces were natural and realistic, aiding them in more effectively accomplishing HRI tasks. Originality/value This study introduces a novel noncontact force feedback approach for HRI, which can significantly improve the control which, compared to initial force feedback methods, allows the operator’s hand to be unrestricted and provides a significantly larger operational space.