https://youtube.com/shorts/XT_Puz_A5Rk
1. BackgroundChallenges in Meeting Room ManagementModern offices need efficient room operations, but the following issues are common:
- Ghost bookings: Rooms are reserved but not actually used
- Opaque utilization: Hard to understand true usage
- Manual inefficiency: Visual checks and manual reservation adjustments
- Real-time friction: Ad-hoc availability checks and extensions are cumbersome
We built RoomBot, leveraging IoT devices and Google Cloud AI. Using M5Stack for image/audio capture plus AI analysis, the system automatically detects, visualizes, and notifies room usage.
2. What is RoomBot?RoomBot is a smart monitoring and conversational system that connects in-room IoT devices (M5Stack) with GCP.
What can it do?- Automated monitoring: Capture images every 5 minutes to detect usage
- AI analysis: Auto-detect headcount with Google Cloud Vision API
- Voice interaction: Handle extension confirmations and notifications via voice
- Real-time visualization: Display live utilization with Firebase
Area
Technology
Device
M5Stack series
Cloud
Google Cloud Platform
AI Services
Vision API / STT / TTS
Database
BigQuery
Frontend
React + TypeScript + Vite
Visualization
Firebase
3. How it Works3.1 Key Featuresπ· Automated Image Analysis- Capture images at 5-minute intervals
- Detect headcount with Vision API and store in BigQuery
- Record reservation data (time, attendees, etc.)
- Detect ghost bookings by matching with images
- Automatically compute utilization rates
For the demo, a custom reservation system is implemented; integration with Google Calendar is planned.
π£ Voice Interaction- Voice notification 5 minutes before meetings
- Voice input to confirm extensions
- Natural Japanese conversation via TTS/STT
- Visualize usage on a dashboard
- Aggregate average headcount in 15-minute intervals
- Analyze trends with daily reports
- M5Stack captures an image and sends it to Cloud Run (blue)
- Cloud Run performs Vision API analysis (blue)
- Results are stored in BigQuery (blue)
- Real-time (15-minute) and daily analyses run (red)
- Visualization via Firebase (gray)
Table Name
Description
meeting_rooms
Basic room information
reservations
Reservation info (clustered by room_id)
room_logs
Headcount detection logs
analysis_results
Daily analysis results
mv_daily_room_stats
Daily statistics view
mv_recent_occupancy_15m
15-minute real-time occupancy stats
3.3 How We Built ItIn-Room Device (M5Stack)- Wi-Fi connectivity (primary/backup)
- ID token authentication
- Image capture and voice notifications
- Voice-based extension confirmation (mic/speaker)
# Representative endpoints
/upload # Image upload
/messages # Retrieve notification messages
/response # Process voice responses
/reservations # Manage reservation data
/rooms # Manage room information
/analysis # Fetch analysis results
Frontend (Firebase/React)- Dashboard views
- UI for managing reservations and utilization
- Analysis result views
- Headcount flow (Device β AI analysis)
- 5-minute pre-meeting voice notification
- Voice-based extension confirmation
- Analytics in Firebase
- UI layout and reservation screen walkthrough
RoomBot combines IoT and AI to make meeting room management simpler and clearer.
β What It EnablesDetect ghost bookingsCompare AI headcount with reservations to find discrepancies.
- Detect ghost bookings Compare AI headcount with reservations to find discrepancies.
- See status in real time Check availability anytime on the dashboardβgreat for ad-hoc meetings.
- See status in real time Check availability anytime on the dashboardβgreat for ad-hoc meetings.
- Handle tasks by voice Extend meetings without opening a laptop; supports natural Japanese.
- Handle tasks by voice Extend meetings without opening a laptop; supports natural Japanese.
- Understand usage patterns Automatic daily summaries help eliminate waste.
Capable on small, low-cost hardware
M5Stack becomes powerful when connected to the cloud.
Capable on small, low-cost hardware
M5Stack becomes powerful when connected to the cloud.
Scales to many rooms
Designed for robustness on Google Cloud.
Simple, maintainable architecture
Minimizes overhead for both builders and users.
π Next StepsProduction rollout with Google Calendar integration
The demo uses a custom system; integration with real schedules is planned.
More personalization
Use face recognition and user info to verify that the reserver is the actual user.
Beyond meeting rooms
Extend to lounges, open spaces, and more across the office.
The RoomBot project began with a simple pain point: βRooms look fully booked, but no oneβs there. If only we could use rooms more conveniently.β By bringing cutting-edge tech to everyone in an accessible way, we aim to make future office work just a bit more comfortable.
Comments