Jido Ants

Jido Ants

A research and simulation project exploring ant colony intelligence using distributed autonomous agents in Elixir. The project implements emergent foraging behavior through individual agents following simple rules, enhanced with machine learning for adaptive pattern optimization.

Project Overview

This project simulates an ant colony where each ant is an autonomous agent operating within a shared environment. The colony exhibits swarm intelligence through:

• Stigmergic communication via pheromone trails that reinforce successful foraging paths
• Direct ant-to-ant communication when agents encounter each other within a 3-unit radius
• Machine learning integration using Axon and Bumblebee for adaptive search pattern optimization
• Terminal-based visualization for real-time observation of emergent colony behaviors

Goals

1. Demonstrate Swarm Intelligence: Show how complex collective behavior emerges from simple individual rules
2. Explore Ant Colony Optimization (ACO): Implement classic ACO principles for path finding and resource allocation
3. Integrate Machine Learning: Enable the colony to learn and adapt its foraging strategies over time
4. Provide Visualization: Build a terminal UI for observing and debugging the simulation

Technology Stack

Component	Library	Purpose
Agent Framework	Jido v2	Autonomous agent framework with state machines, actions, and skills
Language	Elixir ~ 1.18	Concurrent, fault-tolerant runtime (BEAM)
Communication	Phoenix.PubSub	Event-driven messaging between agents and UI
ML / Neural Networks	Axon	Custom neural network models for learning
Pre-trained Models	Bumblebee	Integration of pre-trained models (future)
Numerical Computing	Nx	Tensor operations backing ML components
Terminal UI	term_ui	Elm-Architecture-inspired terminal vi