Ville Kyrki, Lappeenranta University of Technology
Machine learning for abstraction in robotic systems
Since the first autonomous robot, Shakey, was built in late 60's in Stanford, artificial intelligence and robotics have had a long and stormy relationship. Many results of artificial intelligence studies, such as planning algorithms, are essential tools in robotics. However, traditional artificial intelligence approaches have challenges now that robotic systems are appearing more and more in less structured and uncertain environments. The main challenge is to cope with the variability and uncertainty of the world.
The presentation will discuss the use of machine learning approaches in connecting abstract concepts of inference to sensor readings, connecting signals with symbols. Examples of recent research are presented, including topics such as understanding the structure of human motion and abstraction of physical manipulation.
Ville Kyrki is a professor in Computer Science and head of Machine Vision and Pattern Recognition laboratory at the Lappeenranta University of Technology. His research interests lie mainly in intelligent robotic systems and computer vision.