Skip to content

README.md

Latest commit

 

History

History
89 lines (62 loc) · 5.1 KB

README.md

File metadata and controls

89 lines (62 loc) · 5.1 KB

Array Button: Intelligent Workforce & Inventory Ecosystem

Status Architecture Stack

Array Button is an advanced Enterprise Situational Awareness System that transcends traditional inventory tracking. It functions as a high-level operational intelligence layer—effectively a cognitive communication network—that connects physical assets with human workflow in real-time.

Transforming Noise into Intelligence: Array Button turns the chaotic acoustic environment of a job site into structured, actionable business intelligence. Every command spoken is a data point saved, analyzed, and optimized for efficiency.

The Main Agenda

While currently prototyped as a voice-enabled inventory logger, the final vision of Array Button is a comprehensive Workforce Intelligence Platform designed for firms where "deskless" employees operate in complex environments.

Core Capabilities

  • Hyper-Personalized Monitoring:** Moving beyond generic logs, the system builds unique activity profiles for every employee (e.g., "Shreyas is 40% ahead of schedule on steel framing").
  • Automated Progress Reporting:** Eliminates manual status meetings. The system listens to ongoing tasks and auto-generates End-of-Day (EoD) Reports, flagging bottlenecks instantly.
  • Live Status Tracking:** Provides managers with a real-time "Pulse" of the floor knowing exactly who is working on what, where they are, and their resource needs without a single phone call.
  • Predictive Resource Allocation:** Analyzes workforce patterns to predict inventory shortages before they stop production.

System Architectures

Array Button is designed to be hardware-agnostic, supporting two distinct architectural approaches depending on the deployment environment: Distributed AIoT (Current) and Heavy Edge (Alternative/Future).

Architecture A: Distributed AIoT (The "Hive Mind")

  • Current Implementation (PoC)
  • Philosophy: "Dumb Ears, Smart Brain." Low-cost nodes stream raw audio to a central server for heavy processing.
  • Hardware: * Nodes: ESP32 (DevKit V1) & ESP32-C3 (SuperMini).
    • Microphones: I2S Mics (INMP441 / ICS-43434).
    • Server: MacBook M-Series / PC (Python Backend).
  • Software Stack:
    • Firmware: C++ (Arduino), WebSockets.
    • AI Pipeline: Vosk (Wake Word) → OpenAI Whisper (STT) → RapidFuzz (Logic).
    • Latency: < 200ms Wake Word response.
  • Pros: Extremely low cost per node ($5), centralized updates, high-accuracy transcription via server-grade GPUs.

Architecture B: Heavy Edge (The "Autonomous Unit")

  • Alternative / High-Reliability Setup
  • Philosophy: "Smart Edge." Each node is a self-contained computer capable of processing voice locally without a central server.
  • Hardware:
    • Core: Raspberry Pi Zero 2 W.
    • Audio: INMP441 I2S Microphone.
    • Thermal: Active Cooling (Fan + Heatsink) for sustained load.
  • Software Stack:
    • AI Pipeline: Vosk Medium Model running locally on-device.
    • Connectivity: WiFi (Current) → LoRa (Future).
  • Future Integration: This architecture is designed for sites without WiFi coverage. Future iterations will integrate LoRaWAN transceivers to transmit processed text data over kilometers to a central gateway, enabling true off-grid situational awareness.

Tech Stack & Features

Component Distributed AIoT (ESP32) Heavy Edge (Pi Zero 2 W)
Processor Xtensa® Dual-Core 32-bit LX6 Quad-core 64-bit ARM Cortex-A53
Wake Word Vosk Small (Server-side) Vosk Small/Medium (On-Device)
Transcription OpenAI Whisper (High Fidelity) Vosk / PocketSphinx (Edge Optimized)
Connectivity WebSockets (WiFi) WiFi / LoRa (Planned)
Power 0.2 Watts 1-2 Watts
Use Case Indoor / WiFi-covered Zones Remote / Large-scale Sites

Roadmap: From Prototype to Product

The current implementation serves as a Proof of Concept (PoC) validating the distributed acoustic architecture. The roadmap transitions this into a fully integrated Industrial IoT (IIoT) Product.

  1. Contextual AI (Gemini 1.5 Pro): Injecting logs into LLMs to generate narrative reports (e.g., "Why is cement usage 20% higher today?").
  2. Safety Compliance: Training models to detect acoustic anomalies like distress calls, machinery failure sounds, or falling debris.
  3. Biometric Integration: Combining voice ID with wearable health data for fatigue monitoring.
  4. LoRa Mesh Network: Implementing the Architecture B communication layer for multi-kilometer range on construction sites.

Contributors

  • Shreyas Rai
  • Likith K
  • Manoj

Array Button is not just a tool; it is the operating system for the deskless workforce.

Pi.Edge.device.mp4