Reinforcement Learning-Based Admittance Control for Physical Human–Robot Interaction With Output Constraints

Focused on the scientific issues of collision avoidance and compliant operation of physical human-robot interaction (pHRI) systems, this paper proposes a reinforcement learning (RL) strategy based on admittance control to achieve compliant collision avoidance and accurate trajectory tracking of pHRI. Firstly, a differentiable reference trajectory is generated using a soft saturation function with an admittance model. Subsequently, a reinforcement learning strategy based on an actor-critic structure is implemented to address dynamic uncertainty and enhance tracking performance and compliance. Different from existing studies, a reinforcement learning admittance controller containing an integral barrier Lyapunov function (IBLF) is constructed to attain accurate tracking while ensuring that the end-effector achieves the position constraints. Lyapunov stability theory is employed to proof that all states of the closed-loop system remain semiglobally uniformly ultimately bounded (SGUUB). Finally, a suite of tests on Baxter robot experimental platform have been conducted to validate the superiority of the proposed algorithm compared with adaptive impedance control and conventional admittance control. Note to Practitioners—This paper is motivated by solving the safety problem of interaction between collaborative robots and operators, the proposed method is applicable to service industries and industrial sectors such as medical rehabilitation and collaborative assembly. In current human-robot interaction methods, robots typically operate without positional constraints. This absence of restrictions can lead to potential collisions with obstacles or other participants during collaborative tasks between humans and intelligent robots, raising significant safety concerns. This paper presents a novel control method for human-robot interaction. Firstly, admittance control is employed as a foundational approach to enhance the flexibility of human-robot interactions. Subsequently, a RL strategy is implemented to model the uncertainties in the robotic system and to improve tracking performance during interactions. Lastly, a differentiable soft saturation function is utilized for trajectory shaping, and the output of the physical human-robot interaction (pHRI) system is constrained through the integral barrier Lyapunov function (IBLF) to prevent collisions by regulating the robot’s position. Preliminary experiments on the human-robot interaction platform demonstrate that the proposed method is viable, and the controller has application value in many fields. The proposed controller can be applied in industrial robot assembly tasks to ensure that the industrial robot operates within the designated task area. In certain industrial applications, end-effectors of robots are equipped with hazardous or delicate tools. When the end-effector surpasses the defined task area, it may pose a potential risk of injury to both the human operator and the robotic system. Controller can also be utilized in rehabilitation robots to prevent harm to patients by restricting their movements. In some rehabilitation tasks, the controller parameters could be adjusted in accordance with the patient’s feedback to adapt the autonomy of the rehabilitation task. In future research, we will design control for more complex human-robot interactions, such as human-robot collaborative handling and exoskeleton robots.