Cable Assembly Based on Robot Manipulation and Control

Robots play important roles in improving the productivity and efficiency in modern product assembly lines. However, most objects involved in the assembly tasks are rigid. When the objects are flexible and deformable, it is challenging to manipulate them and use them as feedback to control the motion of robots. This paper proposes an assembly framework to first manipulate deformable cables to desired shapes, then to move the robot to target assembly locations to achieve the assembly tasks. The manipulation approach aims to learn the relationship between cable deformation and robot actions based on a convolutional neural network. A vision-based control method is also developed to move the robot to the assembly locations. Experiments are conducted to validate the performance of the proposed framework.