EV-Bharat: Intelligent Electric Vehicle Ecosystem

EV-Bharat is a state-of-the-art, AI-integrated platform engineered to streamline Electric Vehicle (EV) adoption and management in the Indian market. It bridges the gap between low-level vehicle hardware (ESP32/BMS) and high-level predictive intelligence.

---

🏗️ System Architecture

graph TD
    subgraph Hardware Layer
        A[BMS & ESP32 Sensors] -->|Serial/Bluetooth| B(Hardware Bridges)
    end

    subgraph Backend Services
        B -->|WebSockets| C{Central Hub}
        C --> D[Battery Intelligence Engine]
        C --> E[Driver Safety Engine]
        C --> F[RAG Advisory Bot]
        C --> G[YOLO Detection Service]
    end

    subgraph AI Engines
        D -->|Regression| H[SOC/SOH/RUL Predictions]
        E -->|Computer Vision| I[MediaPipe Drowsiness Detection]
        F -->|Gemini LLM| J[Compliance Advisory]
        G -->|YOLOv3| K[Object Tracking]
    end

    subgraph Frontend
        C -->|REST/WS| L[Intelligent Dashboard]
    end

Loading

---

🌐 Network & Service Map

EV-Bharat runs as a microservice ecosystem. Below is the internal port mapping:

Service	Port	Primary Purpose	Check URL
Frontend (Vite)	5173	Core UI & Analytics Dashboard	http://localhost:5173
Node Backend API	3001	User Auth, Profiles, App Data	http://localhost:3001/api/health
RAG Backend	8001	Indian EV Compliance Chatbot	http://localhost:8001/health
ESP32 Bridge	8002	Bluetooth/Serial over WebSocket	http://localhost:8002/status
ML Backend	8000	Battery Analytics + Drowsiness	http://localhost:8000/health
YOLO Service	8010	Real-time Object Detection	http://localhost:8010/health

---

🧠 Machine Learning Model Stack

1. Battery Analytics (Regression)

• Purpose: Predicts SOC (92%), SOH (90%), and RUL (85%).
• Models: RandomForestRegressor, GradientBoostingRegressor.
• Key Files: SOC_model.pkl, SOH_model.pkl, RUL_model.pkl.
• Path: backend/ML/battery_models/

2. Driver Drowsiness (Vision)

• Purpose: Safety monitoring using EAR (Eye Aspect Ratio) and PERCLOS.
• Models: MediaPipe FaceMesh + drowsiness_model.h5 (Keras).
• Path: backend/ML/drowziness/

3. Object Detection (YOLOv3)

• Purpose: Real-time stream analysis from vehicle cameras (ESP32-CAM).
• Files: yolov3.cfg, yolov3.weights, coco.names.
• Path: camera_service/

4. RAG Compliance Bot (NLP)

• Architecture: Retrieval-Augmented Generation.
• Components: FAISS Vector Store + Google Gemini 1.5 Flash/Pro.
• Path: backend/EVBharatBot/

---

🔧 Installation & Startup

I. Prerequisites

# Ubuntu/Debian System Dependencies
sudo apt-get install -y libbluetooth-dev python3-dev libgl1 libglib2.0-0 tesseract-ocr poppler-utils

II. Installation

# 1. Install Node Dependencies
npm install && cd backend && npm install

# 2. Install Python Dependencies
pip install -r backend/requirements_bluetooth.txt
pip install -r backend/ML/requirements.txt
pip install -r backend/EVBharatBot/requirements.txt

III. Startup (Multi-Terminal)

• Terminal A (Frontend): npm run dev
• Terminal B (Node API): cd backend && npm run dev
• Terminal C (RAG Bot): cd backend/EVBharatBot && ./start_rag_backend.sh
• Terminal D (ML Engine): cd backend && python run_ml_backend.py
• Terminal E (Bridge): cd backend && ./start_bluetooth_bridge.sh

---

🛠️ Troubleshooting & Fixes

Issue	Cause	Fix
Address already in use	Port conflict (usually 8000)	lsof -i :8000 -> Kill PID or run YOLO on 8010.
Bluetooth Bridge fails	Missing pybluez or system libs	sudo apt-get install libbluetooth-dev
Login fails	Node backend (3001) is offline	Start backend service in backend/ folder.
RAG fails	Missing GOOGLE_API_KEY	Add key to .env in backend/EVBharatBot/.
Camera unavailable	No webcam access	Check permissions or use v4l2-ctl to verify /dev/video0.

---

📊 Evaluation Summary

• Unified Ecosystem: Seamlessly combines Hardware, AI, and Policy Advisory.
• Edge Ready: Optimized Python bridges for low-power ESP32 hardware.
• Intelligent Proxy: Vite dev server routes /api, /api/rag, and /api/battery to correct microservices automatically.
• Robustness: Integrated fallback to Local Hash Embeddings ensures the chatbot works even without a cloud connection.

---

© 2026 EV-Bharat Team | Advancing Indian Electric Mobility through AI-Driven Excellence.