SAVINA - Humanitarian Logistics AIoT System

AIoT & Humanitarian Logistics system for food distribution and relief: ESP32-CAM with Computer Vision for aerial human detection, K-Means clustering algorithm for resource allocation based on household data, priority, and location, and RFID for warehouse inventory management.

Project Overview

SAVINA is a humanitarian logistics initiative leveraging AI and IoT technologies to optimize resource distribution in Vietnam, spanning hardware (edge AIoT) to software, databases, and ML algorithms:

Designed custom PCB/PCBA with ESP32-CAM for embedded Edge AI computer vision on drones to detect people from the air.

Implemented RFID, temperature-humidity sensors, and ESP32/ESP8266 MCUs as a demo system for inbound, inventory, and outbound warehouse management.

Developed a software stack (Flask, Redis, PostgreSQL) for citizen and admin access, integrating K-Means clustering to allocate relief supplies more effectively, with a pilot case in Ho Chi Minh City.

Role

Firmware | Edge AI | Software Developer
Team SAVINA

Award

2nd Runner-up · HumanLog Hackathon 2025

Product Scope

Hardware → Firmware → Software

System Flow

System Flow & Connectivity

Sensor Layer → Edge MCU:
- IR + Loadcell (HX711) → EFR32xG26 (Sensors) via GPIO interrupts
- SI7021 (Temp/Humidity) + Mic I²S → EFR32xG26 (Edge Drop Detection AI) via GPIO/I²S
Edge AI Processing:
- EFR32xG26 (sound/drop AI) → UDP results
- EFR32xG24 (Perfume Suggestion DL) → LEDs via GPIO + UART
Connectivity Backbone:
- All EFR32 boards on Thread protocol
- MG21 bridges Thread → Raspberry Pi 4
Embedded PC Layer (Raspberry Pi 4):
- Runs Customer Classification AI
- LCD Touch (iPad) user interface
- NLP Transformer for text understanding
Backend & Admin Management:
- Raspberry Pi ↔ Backend over LAN
- Pipeline: MQTT/Redis ↔ Flask ↔ WebSocket ↔ Frontend
- PostgreSQL + Admin Website

S.C.E.N.T Shelf BOM

Smart Shelf Summary

MCU/Embedded PC: EFR32xG26, EFR32xG24, Raspberry Pi 4, EFRMG21
Sensors: HX711 load cells, IR, PIR, Si7021
LEDs: 10 product highlights + 3m decorative strip
Connectivity: Wi-Fi, BLE/Thread, UART/EN0
Power: 220V → multiple 5V adapters
Capacity: Up to 13 SKUs (3 loadcell + 10 IR)
Constraints: 5 kg max/loadcell, PIR 0.5–2 m

Pros

Scalable, flexible, energy-efficient
IoT-ready with real-time connectivity
Multi-sensor fusion for reliable detection
Engaging UX with LEDs for highlight & decoration

SAVINA - Humanitarian Logistics AIoT System

Project Overview

Certifications

System Architecture

Product Stacks

Firmware (EFR32)

System Flow

System Flow & Connectivity

S.C.E.N.T Shelf BOM

Smart Shelf Summary

Pros

Gallery & Artifacts

Downloads & Links

Contact