C++23 implementation of odNEAT: An Algorithm for Decentralised Online Evolution of Robotic Controllers.
Each virtual robot runs the odNEAT algorithm independently with:
- Local timestamp innovation clocks
- Speciated internal population
- Tabu filter
- Broadcast policy
- Maturation guard
A headless e-puck mini-simulator for aggregation, navigation and phototaxis is included.
Ninja is the default generator. If Ninja is unavailable, use the release-make fallback preset (requires Unix make).
Both presets generate compile_commands.json.
cmake --preset release
cmake --build --preset release
ctest --preset release --output-on-failureFallback:
cmake --preset release-make
cmake --build --preset release-make
ctest --preset release-make --output-on-failure./build/run_experiment --task aggregation --cycles 2000 --seed 42
./build/fault_injection --task navigation --cycles 3000
./build/ablation full
./build/minimal_embedConsume as a static library through the public API (include/odneat.h umbrella).
add_subdirectory(odNEAT)
target_link_libraries(my_app PRIVATE odneat::odneat)Installed use:
find_package(odneat 1.0 REQUIRED)
target_link_libraries(my_app PRIVATE odneat::odneat)cmake --preset release
cmake --build --preset release
cmake --install build --prefix /path/to/installMinimal embed (examples/minimal_embed.cpp):
#include "odneat.h"
odneat::OdneatParams evolution_params{};
odneat::OdneatAgent agent(0, 4, 2, 0.0, 100.0, 50.0, 0.0, evolution_params, 42ULL);
agent.executeControlCycle(sensor_inputs, energy_delta, received_genomes);
std::string checkpoint_json{};
std::string error_message{};
odneat::toCheckpointJson(agent, checkpoint_json, &error_message);
odneat::fromCheckpointJson(checkpoint_json, restored_agent, &error_message);JSON checkpoints are the language-agnostic interchange boundary (see
docs/checkpoint_format.md): full-agent resume is bit-identical including the
mt19937_64 stream, populations, tabu history, energy, fitness, clock and
switches. Network activations reset on load; species recompute deterministically.
This repository is an independent implementation of the following research papers:
Silva, F., Urbano, P., Correia, L., & Christensen, A. L. (2015). odNEAT: An Algorithm for Decentralised Online Evolution of Robotic Controllers. Evolutionary Computation, 23(3), 421–449. https://doi.org/10.1162/EVCO_a_00141
@article{silva2015odneat,
author = {Silva, Fernando and Urbano, Paulo and Correia, Lu{\'\i}s and Christensen, Anders Lyhne},
title = {odNEAT: An Algorithm for Decentralised Online Evolution of Robotic Controllers},
journal = {Evolutionary Computation},
volume = {23},
number = {3},
pages = {421--449},
year = {2015},
month = sep,
publisher = {MIT Press},
doi = {10.1162/EVCO_a_00141},
url = {https://doi.org/10.1162/EVCO_a_00141}
}Stanley, K. O., & Miikkulainen, R. (2002). Evolving Neural Networks Through Augmenting Topologies. Evolutionary Computation, 10(2), 99–127. https://doi.org/10.1162/106365602320169811
@article{stanley2002neat,
author = {Stanley, Kenneth O. and Miikkulainen, Risto},
title = {Evolving Neural Networks Through Augmenting Topologies},
journal = {Evolutionary Computation},
volume = {10},
number = {2},
pages = {99--127},
year = {2002},
month = jun,
publisher = {MIT Press},
doi = {10.1162/106365602320169811},
url = {https://doi.org/10.1162/106365602320169811}
}