Skip to content

Overview

Probabilistic Movement Primitives (ProMPs) offer a powerful framework for robot skill learning by representing movements as distributions of trajectories.

In this part of the library we use neural network to create Probabilistic Movement Primitives, thus the name Deep Pro MPs.

Structure

Uml structure of the library

The main part is the class DeepProMP it implements the core training algorithm. It used a DecoderDeepProMP and a EncoderDeepProMP.