Over the last decade, a lot of research has been focused on human motion imitation. Whereas there are lots of methods available how to classify and synthesize those movements, only a few of them are dealing with real life problems in free space or constrained environments. In this context, based only on a few samples a proper movement has to be selected being able to fulfil the desired task while considering constraints and avoiding obstacles.

We are confident that applying a technique called elastic strips (see reference) in combination with basic geometric transformation and a random search algorithm to existing human movement descriptors, a fast and powerful tool being able to generalize human movements can be created.

For a bachelor thesis, the practical part should be executed on a 2D linear axis, for a master thesis, it is intended to be tested on two anthropomorphic arms with 7 DoF each and extended with a learning algorithm in the feedback loop for better overall performan