Cables are widely used in households, healthcare, the automobile industry, and the deep sea/space. However, cable manipulation is still mostly performed manually. Cables are flexible and more fluid in their movement making them difficult to operate by robotic systems. This project is designed to develop robust algorithms for robotic cable manipulation by leveraging a tactile-enhanced visual perception system.

 

In this work, visuotactile sensing is introduced to the traditional vision sensing system for robust robotic cable manipulation. This sensor provides local state estimation of cables in high resolution with high accuracy. This is fundamentally distinct from the traditional approaches. Typically vision sensors are used and are only able to provide global state estimation if they do not suffer from occlusions. In addition, interpretation from only these sensors produces a heavy computation load and can be prone to inaccuracy.

 

This project is built upon She's research team's previous work on cable following and extends to distinct primitives such as cable routing and cable insertion with various constraints. The work will greatly improve the capability of autonomous cable manipulation in various applications, such as the automobile manufacturing industry. Professor Yu She states, "We hope the proposed work would improve the capabilities of robots to do the repetitive, dull, or unsafe tasks for humans, leaving more rewarding and interesting work for the people.