Projects

Here is a list of most of the projects I've worked on. My Twitter showcases additional projects not included here. I also have a YouTube channel with demos/timelapses of several of these projects.

The code to all of these projects can be found on my Github :)

AI learns to play Snake in the Terminal

Neural Network learns to play the snake game in the terminal using Rust and Ratatui.

It uses neuro-evolution to train to grow longer while avoiding death. The AI is capable of solving a 15x15 grid in just about 10-15 minutes.

Code - github.com/bones-ai/rust-snake-ai-ratatui

Genetic Algorithm simulation in C

This is a simple genetic algorithm simulation written in C and Raylib.

The blue rockets are trying to find a path to the green target on the right. Initially they don't know how to navigate the environment, but in a few generations they are able to learn how to get to the target reliably by taking the right turns

Code - github.com/bones-ai/c-smart-rockets

Snake AI (1st Iteration)

This was my 1st attempt at building a snake AI, it wasn't able to solve the game fully.

• Code
• Explanation and Timelapse video

Ant Colony Simulation

An ant colony simulation where ants search for food sources near them and bring the food back to their colony.

They use a mechanism called pheromone trails to navigate around the world, collectively working together to find the food source and return home. It uses a k-d tree to efficiently resolve collisions.

• Video explaining how the simulation works
• Timelapse of the ants working together
• Neon Ants
• Code

2D Self-Driving Car AI

AI learns to drive a car in the classic retro game Road Fighter.

Several cars, each controlled by a neural network, are released into the environment. These neural networks perceive their surroundings using multiple raycasts emitted from the cars. The goal of the simulation is to train a network that can successfully navigate and complete the entire level.

Built in Rust using the Bevy game engine.

• Explanation + Timelapse
• Devlog
• Code

Simulating Evolution in an Ecosystem

An ecosystem simulation of boids learning and adapting to the environment around them.

The boids are controlled by a Craig Reynolds steering behaviours. The simulation explores natural selection, where boids evolve to eat food, avoid poison and predators.

• Part 1
• Part 2 (with predators)
• Code

Training an AI to precisely shoot at a target

This is an evolutionary simulation of AI agents learning to shoot at things around them. The AI was able to learn the basic behaviour of shooting at its target, however I wasn't able to train it to learn advanced strategies like avoiding obstacles and moving away from targets etc.

I'm hoping to come back to this project in the future :)

• Part 1
• Part 2
• Code

Escape AI

AI learns to escape a room.

An AI agent is locked in a room and must find a way out, there's a key that unlocks the door. The AI uses genetic algorithm and reinforcement learning to learn how to navigate around the room, find the key and open the door.

The AI used here was pretty primitive. I look forward to solving this problem again in the future.

• AI learns to escape 5 rooms
• Code

AI learns to play Flappy Bird

This was the first ever AI I built, it uses a simple neural network and a genetic algorithm to learn how to play Flappy Bird.

• Timelapse
• Code