Model Learning for Reactive Motion Planning in Assisting Robots

J. Vain (Estonia) and F. Miyawaki (Japan)


Assisting robots, learning, extended timed automata


We discuss the concept of human adaptive assisting robots and address the issues of model based reactive planning and model learning. We introduce the learning architecture showing interaction between low-level motion trajectory segmentation and high-level symbolic learning components. We propose an algorithm for passive learning from the behaviour of human operator. We demonstrate that the automatic compilation of extended timed automata model in the learning process can be parametrically guided that allows to generate models of different level of abstraction and profile.

Important Links:

Go Back