Head of Department: Dr. Freek Stulp
The Cognitive Robotics department investigates the representation of robot skills and semantic knowledge, as well as methods that enable robots to acquire such skills and knowledge through learning. This is essential to achieve the adaptability and autonomy that robots require to accomplish tasks in unknown and unstructured environments, be it on Mars or in your own living room.
An important topic in the department is developing learning methods for control, motion planning, and task planning. To this end, we use machine learning techniques such as imitation learning – learning from demonstrations – and reinforcement learning – learning from rewards. Furthermore, semantic planning and artificial intelligence are deployed to combine and chain individual skills so that the robot is capable of solving abstract problems autonomously and over longer periods of time.
Human-robot interaction will have a large impact on spaceflight, logistics, manufacturing, but also on our day-to-day lives. Intuitive interaction between humans and robots requires safe control and motion planning, as well as procedures for programming and operating robots that are also accessible for non-specialists. One objective of cognitive human-robot interaction is to analyse biological measurement data, such as electromyograms and skin conductance responses, to recognise and even anticipate human intentions.