Reproduce “Eco-evolutionary Dynamics of Non-episodic Neuroevolution in Large Multi-agent Environments” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Gautier Hamon, Eleni Nisioti, Clément Moulin-Frier
This work introduces a non-episodic neuroevolutionary framework for large multi-agent environments, demonstrating that physiology-driven selection in a continuous setting leads to emergent eco-evolutionary dynamics and the evolution of diverse, adaptively foraging agent populations. The framework successfully runs simulations with thousands of agents over 1 million timesteps, revealing predator-prey oscillations and the emergence of sustainable behaviors.
Key insights
- Eco-evolutionary dynamics, including Lotka-Volterra-like predator-prey oscillations and initial 'near tragedy-of-the-commons' scenarios, naturally emerge from non-episodic, physiology-driven multi-agent neuroevolution.
- A minimal criterion selection mechanism, based on internal energy levels, is sufficient to drive the evolution of complex adaptive behaviors and maintain phenotypic diversity in agent populations.
- Intra-life adaptation, enabled by recurrent neural networks (LSTMs), allows agents to adjust their foraging 'greediness' in response to real-time resource density, contributing to sustainable resource management.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.