º£½ÇÉçÇø

Imperiaus

The Adaptive & Intelligent Robotics Lab (AIRL) has now merged with other leading AI labs at Oxford, UCL and Imperiato form , the British Open-ended Learning and Discovery Lab, a national AI research lab. Our research continues within BOLD’s Embodied Learning pillar. Visit to find out more about our research, people and opportunities.

Our work develops the algorithmic foundations that allow physical robots to learn, adapt and discover new skills on their own. Robots could deliver tremendous benefits to society, but today they mostly operate in well-controlled environments such as factories and warehouses. One reason is that engineers cannot anticipate every situation a robot will face in the real world. Another is that the data-hungry methods behind modern AI transfer poorly to physical systems with limited data and onboard compute.

We develop learning algorithms that enable robots to autonomously adapt to new environments, new tasks and even mechanical damage. Our research combines Reinforcement Learning, Foundation Models, Agentic Systems, Quality-Diversity and Evolutionary Algorithms, World Models, Unsupervised Learning and Open-Ended Learning, with legged robots as our main testbed.