Robotic perception and learning

Academic lead: Professor Rustam Stolkin, Dr Naresh Marturi, Professor Andrea Cavallaro, Dr Miles Hansard, Professor Gerhard Neumann
Partner lead: University of Birmingham, Queen Mary London, University of Lincoln

Work package five will develop novel perception methods, combined with learning techniques to enable robust autonomous operations in variable and unstructured nuclear environments, where sensing may also be impaired by clutter, challenging lighting conditions or darkness, dust, and murky underwater scenes as well as radiation noise and unknown effects of thermal, chemical and mechanical processes. WP5 will work closely with WP4 (novel sensors) to leverage the capabilities of new sensors specifically developed and tailored for nuclear environments; WP7 (grasping and manipulation) to jointly develop algorithms for active gathering of information and closed-loop visually enabled manipulation and grasping; WP6 (mobility and navigation) to co-develop optimal methods for 3D characterisation of nuclear environments; and WP8-9 where cognitive understanding of scenes is essential for variable autonomy, shared control and augmented reality.