Grasping and manipulation

Lead academic: Professor Gerhard Neumann, Professor Rustam Stolkin, Dr Naresh Marturi, Professor Kaspar Althoefer, Dr Mike Mistry 
Partner Lead: University of Lincoln, University of Birmingham, Queen Mary University London, University of Edinburgh

Very complex manipulation capabilities are needed in nuclear robotics. For example, the use-case scenarios of waste handling or cell decommissioning. In these use-cases, the robot must forcefully interact with objects and environment (pushing, cutting, shearing, and grinding – in addition to more conventional pick-and-place actions). These interactions pose challenges to both low and high levels of robotic control, such as which object should be manipulated first, or how should an object be cut. These interactions are so complex that skill acquisition needs to be data-driven, i-e., the robot improved its skills by learning from demonstrations and self-improvement. Data-driven algorithms will be developed in close collaboration with WP8 (HRI interfaces) and WP9 (shared control) to facilitate visualisation of the robots’ plans and interventions from the human operators. Data-driven manipulation algorithms will be integrated in multiple robot platforms, including BROKK hydraulic robots at University of Lancaster, the KUKA industrial robot test-rigs at NNL Ltd Workington and NNL Ltd Preston, the bespoke nuclear manipulators at James Fisher Nuclear and Althoefer novel inflatable manipulators using sensors from WP4.