Overview
AntStack presents a multilevel framework for modeling ant colony organization, from molecular and neural processes at the individual scale through interaction networks to colony-level behavioral patterns. The framework provides a structured approach to integrating diverse biological data across scales in social insect research.
Overview source: Curated paper metadata.
Methods and contributions
Read the source for the full argument, qualifications, and evidence.
Findings and contributions
- The paper presents the Ant Stack, a compact modular framework (AntBody, AntBrain, AntMind) intended to emulate an ant from physics to cognition.
- Stated contributions include an executable I/O contract between layers, a compact neural pipeline with local plasticity, and a minimal active-inference agent composing via stigmergy.
- The work specifies (rather than reports results from) evaluation suites for navigation, trail following, task allocation and robustness under noise/adversaries.
- The author states that no high-fidelity ant brain emulation yet closes the loop with a realistic body at interactive rates.
- Stated limitations: no ant connectome (fly/bee templates used), simulation-first scope with embedded real-time operation as future work, and field validation still needed.
Methods
- AntBody: FlyBody MuJoCo simulator adapted to ant morphology — The body layer adapts the FlyBody MuJoCo simulator to ant leg kinematics, exoskeleton and antennae/mandible actuation, portable to PyBullet.
- Explicit Body–Brain I/O contract with fixed rates and SI units — Specifies observations and actions at 100 Hz over a 1 ms physics step, with a latency budget and clock-drift bound.
- AntBrain AL→MB→CX pipeline templated from fly/bee brains — A functional neural abstraction templated from Drosophila/Apis via Virtual Fly Brain, with sparse Kenyon-cell coding, local plasticity and a ring-attractor CX.
- AntMind minimal active-inference generative model — A single-agent generative model over pose, heading, hunger and pheromone expectation, with variational free energy updates and short-horizon policy selection.
- Diffusion–decay pheromone field for stigmergy — Colony coordination via a grid pheromone field with decay λ, diffusion D and reward-linked deposits, which agents follow along the gradient.
Summary sources: Paper metadata and evidence · Extracted source text.
Document downloads
PDFs are hosted here; supporting sources may use pinned archives.
- 2025_AntStack.pdf PDF · 0.59 MiB · extracted-text source
Citation
Citation metadata follows the unified bibliography.
Catalog details and resources
Related works
Ranked by recorded methods, topics, contributors and associated software, with contributor roles and domain-only matches labelled. Metadata overlap is not a citation or a scientific equivalence.