About

Today's robots perform well only in carefully controlled situations, such as the fully automated assembly lines of car manufacturing; instead, future robots must be able to work in less structured spaces, such as those designed for people, so that they can be easily integrated in a larger number of businesses and operations, and even work in collaboration with human worker. However, current robots lack the required intelligence to manipulate objects with a human-like level of dexterity, and therefore cannot be employed successfully in semi-structured or unstructured environments.

Therefore, we are designing a system that allows human users to easily teach to robots dexterous manipulation strategies that could be used in different real-world situations. By using a set of wearable Human-Computer Interfaces, human users can "move" the robot by simply moving their own hands, "see" through the eyes of the robot and "feel" what the robot is touching. The robot is perceiving the external world by artificial tactile sensing and computer vision, and is "guided" by the human to discover how to best manipulate objects based on such perceptions. In a sense, human users "transfer" their intelligence to the robot, and the robot can reuse it in the future to adapt to a variety of different situations.

We are confident that the results generated by this project will boost productivity in a large number of manufacturing processes at different scales, both for large industries and for SMEs (economic impact), and they will improve the working conditions and quality of life of human workers in terms of safety and engagement (social impact).

The project is led by Dr Lorenzo Jamone and developed by the CRISP robotic group. The CRISP group is part of the Advanced Robotics at Queen Mary (ARQ), School of Electronic Engineering and Computer Science (EECS), Queen Mary University of London (QMUL). We are happy to collaborate with: Shadow Robot Company, Google DeepMind, Ocado Technology. The project is funded by the EPSRC UK (grant #EP/S00453X/1), for a duration of 3 years (01/01/2019-31/12/2021).