This repository contains the source code for all the experiments presented in the following publication:
Huizinga J, Clune J. (2018). "Evolving Multimodal Robot Behavior via Many Stepping Stones with the Combinatorial Multi-Objective Evolutionary Algorithm". arXiv:1807.03392.
If you use this software in an academic article, please consider citing:
@article{huizinga2021evolving,
title={Evolving multimodal robot behavior via many stepping stones with the combinatorial multi-objective evolutionary algorithm},
author={Huizinga, Joost and Clune, Jeff},
journal={Evolutionary computation},
pages={1--34}
year={2021}
}
For ease of reproducing the experiments, the code includes the sferes2 framework and all modules that were used.
The robotics experiments rely on the Bullet physics engine. Different version of bullet may work, but we recommend bullet 2.87 compiled with demos and double precision
And configure and build the experiment:
./waf configure --bullet-double-precision
./waf --exp cmoea
./waf --exp modularity
The function and maze domain experiments can be found in:
build/exp/cmoea
The robotics experiments can be found in:
build/exp/modularity
The code in this repository is licensed under the CeCILL License.