Sensorless robotized cable contour following and connector detection

This paper focuses on the identification of the connector of a semi-deformable object (SDLO) (i.e. a deformable element as a cable featuring rigid parts as connectors) to perform insertion and disconnection tasks. We propose a cable tracing skill that allows a robotic arm to follow the unknown contour of a cable held for a small portion by a fixture, and to identify its connector that can be grasped and inserted or disconnected. No prior knowledge of the SDLO's length and shape is required. To slide along the cable, the last joint of the manipulator is made compliant to adapt to the SDLO shape and a recursive least squares algorithm with forgetting factor is exploited to estimate the future local shape of the cable and to speed up the contour following procedure. The detection of the connector is based on the estimated contact force acting on the robot's end-effector while sliding along the SDLO. Remarkably, the method does not require vision, tactile and/or force sensors, it makes use just of a parallel gripper equipped with simple fingertips, and it can be implemented directly in a generic robot programming language. The effectiveness of our strategy is assessed via experimental tests on two different SDLOs in several configurations with connected and unconnected ends. Finally, the skill is also exploited in the robotized assembly of a motorbike braking system.