Skip to content

README.md

Latest commit

 

History

History
160 lines (115 loc) · 5.86 KB

README.md

File metadata and controls

160 lines (115 loc) · 5.86 KB

Carte Diem - Smart Shopping Cart

Abraham Vega, Ava Chang, Felix Lu, Jin Wook Shin, Natalie Do, Tina Ma

{abevega, avachang, felixlu, jinws, donatali, tinma}@umich.edu

EECS 473 F25

Expo Poster

Carte Diem is a add-on module that integrates smart features to regular carts. This device eliminates the need for checkout lines while maintaining anti-theft features. It’s modularity allows stores to easily implement this technology into the existing ecosystems.

README Directories

  • PCB — schematics, layout, and power architecture
  • CAD — mechanical enclosures and mounting hardware
  • cart - UI interfacing and backend servers
  • cart_tracking - frontend and backend server for cart tracking
  • Integration — ESP32 ↔ Raspberry Pi integration and BLE communication

Motivation

Despite the widespread adoption of self-checkout, grocery stores still experience congestion, long wait times, and loss-prevention challenges. Existing smart cart solutions address these issues but come at a cost of $5,000–$10,000 per cart, making large-scale deployment unrealistic for most retailers.

Carte Diem explores a different design space:

  • Retrofit instead of full cart replacement
  • Sensor fusion instead of single-modality detection
  • Real-time embedded verification instead of post-hoc auditing

The goal is to eliminate checkout lines entirely while maintaining store-side trust, scalability, and cost efficiency.

System Overview

image image

Carte Diem consists of two tightly coupled subsystems:

On-Cart Embedded System

  • Custom ESP32-S3 PCB running FreeRTOS
  • Interfaces with all sensors and peripherals
  • Performs real-time sensing, verification, and event generation
  • Communicates with the Raspberry Pi via BLE

On-Cart Compute + Backend

  • Raspberry Pi 4 provides UI, networking, and payment flow
  • Connects to store and Carte Diem servers over WiFi
  • Handles inventory queries, checkout confirmation, and analytics upload

Key Features

Item Detection & Verification

  • Barcode Scanning

    • Primary item identification method
    • Button- or proximity-triggered
    • Fast and reliable for most items

  • RFID Item Verification

    • Detects tagged items (e.g., alcohol, high-value goods)
    • Automatically cross-checks cart contents
    • Flags unscanned additions in real time

  • Weight-Based Verification

    • Four load cells under the cart basket
    • Cross-checks expected vs. measured weight
    • Detects unscanned add/remove events
    • Freezes checkout when inconsistencies are detected

  • Produce Scale

    • Dedicated top-mounted scale
    • Computes price dynamically based on weight

Checkout & Security

  • Checkout & Payment

    • Simulated RFID tap-to-pay
    • QR-code payment flow (Square API)
    • Digital receipt generation

  • GPS-Based Security

    • Store-defined cart boundaries
    • Alarm triggered if cart exits allowed area

Analytics & Store Insights

  • Cart Tracking
    • RFID-based indoor tracking (no computer vision)
    • Generates cart paths and heatmaps
    • Aggregates multi-session shopping behavior

Hardware Architecture

  • Custom ESP32-S3 PCB
    • UART: barcode scanner, RFID modules
    • SPI: payment RFID
    • I²C: IMU, proximity sensor
    • ADC + GPIO: load cells
    • BLE: ESP32 ↔ Raspberry Pi
  • Raspberry Pi 4 with touchscreen display
  • Dedicated 12 V LiFePO₄ battery
  • Multi-rail power system (12 V, 9 V, 5 V, 3.3 V, 1.8 V)
  • 3D-printed enclosures designed for durability and weather resistance

The system is designed to operate 15+ hours continuously, matching a full day of store operation.

Software Architecture

  • FreeRTOS-based firmware

    • Periodic tasks for sensing and monitoring
    • Event-driven verification triggered by weight changes
    • ISR-to-task deferral for responsiveness and power efficiency

  • BLE Communication Protocol

    • Structured commands for barcode scans, weight readings, RFID bursts, and cart tracking logs

  • Full-Stack Web Backend

    • Store database and inventory management
    • Cart registration and session tracking
    • Heatmap and path visualization
    • GPS boundary configuration interface

Repository Structure

.
├── CAD/               # Mechanical CAD files for enclosures and mounts
├── PCB/               # Schematics, PCB layouts, and fabrication files
├── cart/              # On-cart UI (Raspberry Pi touchscreen interface)
├── integration/       # ESP32 ↔ Raspberry Pi integration and BLE logic
├── LICENSE            # GPL-3.0 license
└── README.md

Results

  • Fully functional prototype demonstrated at the EECS Design Expo
  • Reliable real-time detection of item add/remove events using sensor fusion
  • Weight-based verification successfully detected unscanned items
  • RFID-based cart tracking enabled path reconstruction and heatmap generation
  • End-to-end checkout flow validated, including UI, payment simulation, and receipt generation
  • Power system met full-day operation targets under worst-case load conditions

Photo Dump

image image image