🚘 CrewRide: Smarter Carpooling with AI

CrewRide is an AI-enhanced corporate carpooling platform that matches employees for shared rides based on behavioral compatibility, commuting patterns, and social preferences. The system integrates a backend built with Next.js, a mobile Android app, and machine learning clustering techniques to create more enjoyable and socially attuned carpooling experiences.

---

🌟 Features

• Socially-Aware Ride Matching: Matches co-riders based on personality traits, commuting schedules, and preferences.
• LLM-Powered Ride Preparation: Offers personalized small-talk suggestions using a fine-tuned large language model.
• Real-Time Ride Management: Users can offer, find, and manage carpools via the mobile app.
• Dynamic ETA Mapping: See live routes and driver/rider info on embedded Google Maps.
• Behavior-Based Clustering: Uses Deep Embedded Clustering (DEC) for grouping compatible riders.

---

🗂️ Repository Structure

CrewRide/
├── backend/                # Next.js backend API
│   ├── prisma/             # Prisma ORM schema and migrations
│   ├── public/             # Static assets (e.g., images, icons)
│   ├── src/                # API route handlers, logic, utils
│   ├── .env                # Environment variables
│   ├── next.config.ts      # Next.js config
│   ├── tsconfig.json       # TypeScript config
│   └── package.json        # Backend dependencies and scripts

├── clustering/             # ML-based clustering pipeline
│   └── crewride_dec_pipeline.py  # DEC-based clustering logic

├── database/
│   ├── crewride.db         # SQLite database used for clustering
│   └── crewride_db_gen.py  # Script to generate the DB

├── mobileapp/
│   └── CrewRide/           # Android app source (Java/Kotlin)
│       ├── app/            # Main Android application code
│       ├── gradle/         # Gradle wrapper config
│       ├── build.gradle.kts        # Kotlin DSL build script
│       └── secrets.properties      # API keys or app secrets

├── README.md               # This file
├── .gitignore              # Git ignore rules
└── .gitattributes          # Git text encoding rules

---

⚙️ How It Works

🔁 Ride Matching Flow

1. User Location & Destination Input
2. Backend Search API Filters:
   • Available seats
   • Cluster membership
   • Do-not-match list
   • Departure time window
   • Proximity (max distance & bearing checks)
3. Candidate Prioritization:
   • Preferred-by list
   • Average rating
   • Punctuality score
4. Sorted Results Returned to Client

🤖 Clustering Logic (DEC)

• Applies Deep Embedded Clustering on behavioral and commute preference data from employees
• Outputs cluster ID per employee used in ride matching

🧠 LLM Ride Prep

• Large Language Model provides context-aware conversation starters (e.g., weather, light banter, shared interests)

---

📱 Mobile App (Android)

The Android app (under mobileapp/CrewRide/) features:

• Tabbed UI for upcoming/completed rides
• Ride offering dialog with social preference options
• Map-based view with ETA and participant list
• Direct call/chat/prep buttons for ride coordination

---

🚀 Getting Started

Prerequisites

• Node.js v18+
• Android Studio
• Python 3.9+
• SQLite3

Note: node_modules/ directory in the backend setup has been excluded from this repository view due to its large file size.

Backend Setup

cd backend
npm install
npx prisma generate
npx prisma migrate dev
npm run dev

Clustering Pipeline

cd clustering
python crewride_dec_pipeline.py

Mobile App

Open mobileapp/CrewRide in Android Studio and run the app on an emulator or physical device.

---

🧪 Future Work

• In-the-wild user evaluation and clustering parameter refinement
• Fine-tuning LLMs for carpool etiquette and real-time adaptation
• Designing equitable ride contribution frameworks
• Deriving vehicle design insights from shared mobility data

---

📜 License

GPL-3.0 license. See LICENSE for details.

---