2023-24-project-catalogue

###Self-supervised learning of soft robot body schema using touch and vision

Project ID: 2228bd1188 (You will need this ID for your application)

Research Theme: Artificial Intelligence and Robotics

UCL Lead department: Computer Science

Department Website

Lead Supervisor: Thomas George Thuruthel

Project Summary:

The body schema is an internal representation of the body structure and registration of sensory modalities with respect to this structure. In natural agents, it is vital for perception and control. Robotic systems typically employ mathematical models derived from human knowledge to perform the same function. As robots become more physically complex with numerous sensory modalities, it becomes difficult to develop mathematical models and perform manual sensor calibration. Soft robotic systems, that are gaining high popularity, share the same issues, aggravated also by fabrication variabilities. This project aims to take inspiration from human developmental sciences to develop methods for self-supervised (or weakly supervised) generation of body schema. We will investigate the fusion of visual and tactile information to develop such representations, leveraging on some of the new advancements in deep learning and soft tactile sensor technologies.