Digital Twin Driven One-Step Fastening in Robotic Flexible Printed Circuit Assembly

Nowadays, robots are widely used in various assembly tasks in industry. However, in the assembly task of tiny parts such as mobile phone flexible printed circuit (FPC), it is often necessary to conduct multi-step analysis of multi-modal information such as visual and tactile information to achieve the final fastening of mobile phone FPC. And because the robot has errors in perceiving tactile information, the resulting multi-step fastening strategy is not reliable enough, which makes it difficult for the robot to improve the fastening accuracy and success rate of mobile phone FPC. To this end, this work proposes a digital twin (DT) driven one-step fastening method for mobile phone FPC based on adaptive gradient. The contributions mainly include the following. First, in order to reduce the perception error of real tactile information, we proposed a tactile perception method in a DT environment. In addition to obtaining real tactile data through tactile sensors in a real environment, virtual tactile data is also obtained by simulating the deformation of the soft cable using the material point method (MPM) in a virtual environment. Secondly, in order to achieve the one-step strategy for fastening mobile phone FPC, we proposed an adaptive gradient prediction model based on virtual and real tactile images. By comparing the virtual and real tactile image sequences in the DT environment, adaptive gradient calculation is performed, and the adjustment strategy sequences are input into the LSTM network to predict the adjustment strategy at the fastening moment, thereby achieving fast and efficient fastening of mobile phone FPC, effectively improving the execution efficiency of the robotic arm and the synthetic power of the mobile phone FPC fastening. The experiments were carried out on the actual assembly platform of the Redmi Note 11 mobile phone. With the number of failed trails limited, our approach significantly outperforms vision-only and DT-free approaches.