Reinforcement Learning Car Simulation 🚗🤖

This project demonstrates a Reinforcement Learning agent built in Unity using custom Neural Networks and Genetic Algorithms.
The goal: evolve a self-driving car agent that learns to navigate efficiently.

---

🔑 Features

• Custom Neural Network in Unity (C#) with tunable hidden layers (see NNet.cs).
• Genetic Algorithm Framework (see GeneticManager.cs):
  • Selection, crossover, and mutation operations.
  • Fitness-based repopulation loop to evolve agents.
• Simulation: Cars drive in Unity environment, evolving every generation.
• Results: Top-performing agent achieved by Generation 2, Genome 25.

---

📂 Project Structure

• GeneticManager.cs → Handles population management, fitness evaluation, and reproduction.
• NNet.cs → Implements feed-forward neural network with customizable layers.
• Unity Scene → Contains car prefab and track for training.

---

🛠 Tech Stack

• Unity (C#) for simulation
• Genetic Algorithms for evolution
• Custom Neural Networks for decision-making

---

📊 Visual Demo

Car Simulation Demo

---

🚀 How to Run

1. Clone repo and open in Unity.
2. Attach NNet.cs to car prefab, and GeneticManager.cs to scene manager.
3. Hit Play → Watch cars evolve across generations.

---

📈 Key Takeaways

• Demonstrates neuroevolution techniques in Unity.
• Reinforcement learning without backpropagation, using genetic optimization.
• Framework can be extended to more complex environments.

---

👤 Author

Pradyumn K. Pottapatri
📍 Data Scientist | ML Engineer | NLP, Time Series, Reinforcement Learning, GenAI LinkedIn | GitHub