AI for Robotics
Robots, and autonomous agents in general, have become common place in industrial settings to execute carefully controlled tasks, but their presence in the home and office is not yet quite so well established. Safety issues and a lack of autonomy are the main barriers to this adoption which we believe can be overcome by using artificial intelligence. Our goal is to use AI to make robots a part of our everyday spaces to help and assist us in our everyday lives.
One reason robots lack autonomy is that they simply do not have the ability to acquire and apply ‘common sense knowledge’ to accomplish tasks without being provided with detailed instructions. For example, humans typically don’t need a map to help them leave a room – pinpointing where the door is situated is enough to know where to head for. When we go to pick up a glass of water, we don’t need a detailed physical model describing the conservation of energy or the laws of gravity. In both cases, we apply reasoning and the common sense we’ve garnered from experience.
Similarly, if we know how to pick up a glass at home, we can easily pick up a glass in a restaurant on the other side of the world because we transfer skills to different contexts. We believe that robots also need to be able to transfer acquired knowledge to new contexts or domains just like people, and, ideally without relying too much on human supervision.
As we try to teach robots the capacity for common sense and the ability to adapt to a wide range of contexts, we want them to acquire it from past experience but also by taking appropriate exploratory actions. Being equipped with these capacities will allow robots to operate in real-world environments and perform useful tasks in collaboration with people. To reach this goal, we focus our research on the following areas: navigation, human-robot interaction and physical interaction with the environment.