Representation & Reasoning

Academic Contact: Tony Cohn
Academic Staff: Professor Alan Pearman, Professor Anthony Cohn, Professor Mark Mon-Williams

We are exploring the combination of qualitative reasoning with video-based object detection and robotics to model and recognise everyday activities from the spatio-temporal relationships between people and objects. We are also developing new ways to give meanings to words and text through linking video clips with aligned textual descriptions.

Much of our work depends on representing information in a high-level way to integrate and reason about information from very different sources, including video from cameras; text, images and sounds from the web; numerical measurements from wireless sensors; and other sources of geo-spatial information. We apply semantic technologies using ontologies as a tool to describe the content and organisation of knowledge in different domains.

Underpinning this integrative research, there is fundamental theoretical work on qualitative spatial reasoning, building on our pioneering work on the Region Connection Calculus (RCC); algorithm development for pose estimation, automatic textual alignment and segmentation; research on user modelling and personalisation, and on corpus based language analysis.

To extend the reach and integrative potential of our work, we have recently expanded to include research on robotic manipulation and planning.

Our research has greater impact through working in large multi-disciplinary teams, spanning other engineering disciplines, clinicians, large and small companies and the public-sector. Through these collaborations we are helping to transform construction, transportation, engineering design, security, health care, medical diagnoses, and the maintenance of city infrastructure.