Bestowing our machines with intelligence has been an ambitious goal for scientists since the birth of artificial intelligence as a field in the 1950’s. And as we scientists learn to pace ourselves, the field of artificial intelligence has in the last two decades matured into actually usable techniques for robots. ROBOTNOR conducts research on how robots can automatically improve based on experience. We refer to this as robot learning.
Our research on robots that learn is founded on a desire to create more usable robots for end-users. By making a robot able to learn, we effectively reduce the need for programming the robot. The idea is to enable the robot to learn how to perform its intended task, either by being shown how to perform the task or by being told what its goal is (what to do) instead of how to do it. Subsequently, we would like the robot to experiment (or play) with its environment in order to discover solutions on its own. You may think of this approach as a new kind of user interface for robots.