A Motion Control Approach for Physical Human–Robot-Environment Interaction via Operational Behaviors Inference

Human–robot collaboration systems aim to improve working efficiency and reduce human workload. However, inefficient assimilation of human potential behaviors often leads to increasing human–robot conflicts. In this article, a motion optimization approach integrating behavior inference is presented for physical human–robot-environment interaction (pHREI) tasks to implement assistive behavior. A multistep human behaviors model with long short-term memory (LSTM) mechanism is employed, enabling continuous prediction of the human intention based on robot and environment states. A two-layer control scheme is developed to optimize the manipulator trajectory with the objective of reducing human workload in a model predictive control (MPC) fashion. A series of experiments is conducted to verify the proposed scheme on a wood-sawing task. With the integration of the human and environment model, the proposed control scheme significantly reduces the human workload while eliminating the human–robot conflict simultaneously.