Abstract
Humans move easily through their surroundings due to a powerful and efficient sense of vision. Robots originally lacked this ability to see the shape of the world, but they are now catching up fast.
In this lecture, I will talk about my work on computer vision methods to solve the core `Simultaneous Localisation and Mapping’ (SLAM) problem that robots face in estimating the shape of the world when seen through a mobile camera. I will illustrate progress with live demonstrations of landmark systems.
Improvements in algorithms alongside the exponential rise in computer power is starting to make visual SLAM commercially feasible in small, low-cost devices like domestic robots and smartphones. Meanwhile, academic research in robot vision is edging towards higher level artificial intelligence goals of object understanding and interaction.
Biography
Andrew Davison grew up in Kent and studied physics before becoming hooked by computer vision as a DPhil student in Oxford’s Active Vision Lab. He spent two fantastic years as a post-doc in Japan before returning the UK, and came to Imperial in 2005, always following his interests in real-time localisation and mapping research.
The highlight of his job is working with the tireless and talented members of the Robot Vision Research Group, and he is fortunate to have received extended and unfettered funding and the support of colleagues to follow long-term research goals. He has long-term collaborations with industry to apply vision techniques in robotics and related areas.
Outside of work he enjoys sharing fun and interesting moments, projects and adventures with family and friends, and this year got as far as Paris on a bike.