MORLBENCH Framework

Multiobjective Reinforcement Learning Benchmark Suite

Chair of data processing @ Technical University of Munich

Dominik Meyer meyerd@mytum.de, Johannes Feldmaier johannes.feldmaier@tum.de, Simon Wölzmüller ga35voz@mytum.de

This is the MORLBENCH Framework for algorithm and environment performance testing in multiple objective Reinforcement Learning.

written in Python 2.7

1. After cloning the repository you should run python setup.py develop in parent directory of the framework. (provided you have python 2.7 on your system installed. Otherwise download it(Python Website))

2. You can use every class in morlbench/morl_agents.py and morlbench/morl_problems.py where agents stands for algorithms and problems for the scenarios you want to test in.

3. For first examples using the framework relate to morlbench/scratchpad

To run test type python -m unittest2 test.py in morlbench/test