🧬 COVID Digital Twin & Containment Simulation

A digital twin of an urban environment powered by AI agents and Reinforcement Learning to simulate, analyze, and optimize pandemic containment strategies.

---

🎯 Objectives

• Simulate Realistic Human Agents to Model Urban Dynamics
  Create human-like agents with lifelike demographics, routines, and behaviors that evolve over time, reflecting real-world social dynamics. This helps uncover patterns that drive smarter policy and resource planning during pandemics.

• Train RL Agents for Disease Containment
  Utilize Reinforcement Learning to develop adaptive containment strategies (like dynamic lockdowns) that minimize disease spread while balancing economic and societal impacts.

---

🧠 Overview

This project combines agent-based modeling with reinforcement learning to build a dynamic digital twin of a city (modeled on Los Angeles). Each AI agent simulates real human behavior, allowing the system to evaluate and adapt containment strategies in response to pandemic scenarios.

---

🔑 Key Features

🏙️ Digital Twin of a City

Simulates thousands of realistic human-like agents in a virtual city with detailed spatial geography and movement dynamics based on historical epidemic/pandemic data.

🔁 Chain-of-Thought Driven Routines

Agents follow daily schedules generated using large language models to mimic diverse real-world behaviors and responses under epidemic conditions.

🧪 RL-Based Policy Optimization

Custom OpenAI Gym environment enables RL agents to learn dynamic intervention policies (e.g., selective lockdowns) using algorithms like Proximal Policy Optimization (PPO).

🏥 Predictive Healthcare Insights

• Forecasts shortages in medical equipment, ICU beds, and medicine.
• Simulates the impact of travel restrictions and lockdowns on healthcare logistics.

🧩 Modular Architecture

• Frontend: Interactive dashboard (Next.js) with maps (Leaflet) and charts for visualization.
• Backend: Flask-based API hosting the simulation logic and chain-of-thought reasoning.
• ML Module: Scripts and environments for RL training using stable-baselines3.

---

🧬 How It Works

🔹 Agent-Based Simulation

Each agent is initialized with a profile and dynamic routine. They interact, move, and influence infection dynamics across city neighborhoods.

🔹 Data-Driven Urban Modeling

Tracks time-series data like population density, infection rates, and economic loss. Enables realistic forecasting and planning.

🔹 Reinforcement Learning for Policy Optimization

Encapsulated in a custom Gym environment. The RL agent learns when and where to apply interventions to contain disease spread optimally.

🔹 Visualization & Insights

Real-time maps and dashboards show:

• Agent movements
• Infection spread
• Hospital resource predictions
• RL policy plans

---

🖥️ Tech Stack

---

📷 Gallery

---

🚀 Getting Started

🔧 Prerequisites

• Node.js & npm
• Python 3.8+
• MongoDB

📁 Directory Structure

HealthX/
  ├── backend/        # Flask app, simulation engine, RL environment
  └── frontend/       # Next.js frontend with interactive maps & controls

---

🛠️ Manual Setup

1. Backend

   cd HealthX/backend
   pip install -r requirements.txt
   python app.py

2. Frontend

   cd HealthX/frontend
   npm install
   npm run dev

---

📊 Sample Outputs

• 📌 Interactive Map: Live city simulation with infection and economic overlays.
• 📈 Containment Timeline: RL-generated policy sequence (lockdowns/travel bans).
• 🏥 Hospital Insights: Shortage forecasts and surplus indicators.
• 🧠 Gemini Suggestions: AI-generated recommendations for resource management.

---

📌 Note

This project is a research-oriented prototype and not intended for real-world deployment without further clinical and epidemiological validation.