energy_efficient_flocking

• ants26_replication/ -- current work: the Hebbian ABCD replication of Mahdavi et al.'s ANTS 2026 paper, plus hardware deployment for real Thymio+Pi swarms.
• initial_implementation/ -- the original LJ-force flocking model this project grew out of. Superseded by ants26_replication/ but kept for reference/comparison.
• crazythymio/ -- hardware stack for the next round of hardware experiments: Thymio + Crazyflie board with Lighthouse positioning in place of OptiTrack. Upstream code from https://github.com/fudavd/CrazyThymio (see Acknowledgements); our self-contained deployment of the Hebbian controller on it (own config, custom firmware, launcher, setup/calibration guide) is in crazythymio/lighthouse_deployment/ and crazythymio/firmware/ -- start with crazythymio/lighthouse_deployment/README.md. Independent of the OptiTrack deployment in ants26_replication/hardware_deployment/.
• hardware_transfer_test/ -- evaluates three conditions side by side (the LJ Table-3 rule-based baseline, and two evolved Hebbian genomes -- one from before the hard safety-clamp reflex layer existed, one trained with it active) across n_agents in {2,3,4,5,10,20}: archived per-run metrics.json/video.mp4, plus leadership_metrics.py/leadership_plots.py/paper_figures.py (turn-taking and leadership-network analysis, reproducing each condition's trajectory from its saved genome/rules -- see that directory's own docstrings for the exact reproduction recipe) and overleaf_summary/ (a ready-to-compile LaTeX methodology + results write-up with figures as PDF/SVG/PNG).

ants26_replication/ and initial_implementation/ are standalone Python projects (each its own experiment/ package) that share no imports, only some physics constants happen to have the same values by necessity (see ants26_replication/experiment/config.py's docstring). hardware_transfer_test/'s analysis scripts are the one exception: they import directly from ants26_replication/experiment/ rather than duplicating it.

Acknowledgements

crazythymio/ contains code from fudavd/CrazyThymio (Fuda van Diggelen et al., MIT License, copied unmodified along with its LICENSE), the codebase of Learning online adaptive emergent control for heterogeneous robot swarms. It is used here for the Thymio/Raspberry Pi/Crazyflie-board interface (state estimate from the Lighthouse positioning system, IR-based quadrant sensing, motor commands). The upstream simulation submodule (sim_exp), firmware submodule (CrazyThymio-firmware, from https://github.com/tugayalperen/CrazyThymio-firmware) and the results/ data were not copied; see the upstream repository for those. crazythymio/firmware/app_share_pos_hebbian/ is derived from the firmware's app_share_pos example (GPL-3.0).