Code

I developed this software as part of my studies. Some of it I wrote for a university project, others I wrote for my master thesis.

PyRLtR

This reinforcement learning framework is based on the reinforcement learning framework I build for my masterthesis. It combines multilayer perceptron with the forwards kinematics engine to allow a learner to learn how the reach for a arbitrary target. Hence the acronym Reinforcement Learning to Reach (RLtR).

Currently it supports using reinforcement learning to learn the inverse kinematics of a two DoF robotic arm. It still work in progress and the examples may not work perfectly.

You can find the documentation here.

PyMLP

This small package is my implementation of a Multilayer Perceptron (MLP). It is fairly small and should be easy to use.

Of course, you can also download the Source package and browse the documentation.

Forward Kinematics

I also implemented a small forward kinematics engine which contains all necessary operations for rotationally transformations. It can transform any 3 dimensional position from the shoulder’s coordinate system to the end-effector’s one and back.

Its test class also demonstrates the usage.

Communication Framework

I developed this simple communication framework in Java for a university project in which we wanted several robots to look for and collect coloured barrels. This project, as well as the behaviour depend on additional projects, which I did not develop. But maybe they give someone a base for further development.

The source code is, of course, available for download and you can read the documentation.

Behaviour

I also developed the necessary behaviour, which is very simple for the intended task.

The source code is, of course, also available for download and you can read the documentation.