An augmented reality-assisted interaction approach using deep reinforcement learning and cloud-edge orchestration for user-friendly robot teaching

•An improved deep reinforcement learning algorithm-driven robot motion planning approach is proposed to calculate the optimal path to avoid potential obstacles, thereby reducing the difficulty of robot teaching.•An augmented reality-assisted user-friendly interaction interface is constructed to help operators teach the robot without the limitations of spatial and human factors. Meanwhile, the motion planning trajectories of virtual robots can be integrated into the physical space through AR glasses, which can avoid some potential safety issues before being applied to physical robots.•A cloud-edge orchestration mechanism is designed to link the communication between the industrial AR cloud platform and the edge nodes to provide a smooth and immersive 3D visual robot teaching experience. With the help of this, data-heavy and time-consuming tasks can be fleetly calculated on the cloud platform, whereas interaction tasks with minimal computation can be easily accomplished on the AR glasses. Industrial robots have emerged as pivotal components in the search for intelligent manufacturing equipment that can meet flexible and customized operational needs. Consequently, industrial robots have to frequently use motion planning schemes pre-programmed by operators. Furthermore, traditional robot teaching methods in the human-robot interaction scenario can only be applied in a fixed task environment and therefore lack generalization ability. To address these shortcomings, this research proposes an augmented reality-assisted interaction approach using deep reinforcement learning and cloud-edge orchestration for user-friendly robot teaching. Firstly, the proposed deep reinforcement learning algorithm with the position prediction function is applied for the robot motion planning, which can avoid unnecessary collision attempts during the training process. Subsequently, augmented reality glasses provide a user-friendly interaction interface, allowing both virtual and physical robots to be operated to eliminate the limitations of spatial and human factors. Apart from this, the robot target positions can be set by operators, and the visible trajectory of the calculated path can be integrated into the real scenario by virtue of AR glasses. On top of this, the cloud-edge orchestration links the communication between the industrial AR cloud platform and the edge nodes (e.g., robots and augmented reality glasses). Ultimately, comparative numerical experiments are conducted in an