PROJECT Deep Learning From Demonstrations

AutoLab, UC Berkeley

Code for training an ABB YuMi robot to perform manipulation tasks from observing human demonstrators.

Policy Interface

This package provides a way to collect demonstrations (Kinesthic or Teleoperation) from a robot, train neural network policy primitives and string toward different manipulation tasks.

Installation

Step 1: Install the alan YuMi python interface on the client computer that will communicate with the YuMi:

$ python setup.py develop