Skip to main content

Smart Buildings / Real Estate / PropTech

Digital twin: Sensor units and edge gateways for a smart-building platform

Rapid Circuitry designed custom ESP32-S3 multi-sensor units, NVIDIA Jetson floor gateways with BACnet/Modbus translation, and the data pipeline feeding a digital-twin platform for a 50-story mixed-use tower. The building owner's energy and maintenance results are not published here.

What we did: We designed the multi-sensor units and edge gateways, integrated them with the building's legacy systems over BACnet and Modbus, and built the data pipeline and dashboards on Azure Digital Twins.

Commercial real estate developer14 months

Published Last reviewed:

ESP32-S3
Custom multi-sensor units
BACnet / Modbus
Legacy systems bridged
< 100 ms
Local decision latency
Design target
7 days
Gateway offline buffer
Design target
Illustrative photo: a wireless building sensor in front of a glass office tower at duskIllustrative image

The challenge

A commercial real estate developer managing a 50-story mixed-use tower with offices, retail, and parking facilities needed to modernize their building management systems to reduce operational costs, improve tenant satisfaction, and achieve sustainability certifications.

Energy inefficiency

The building's legacy HVAC, lighting, and elevator systems operated in silos with no intelligent coordination. Peak demand charges and inefficient scheduling resulted in excessive energy waste.

Impact: 38% above benchmark energy consumptionClient-reported baseline

Reactive maintenance

Building systems failed unexpectedly, causing tenant complaints and emergency repairs. The maintenance team had no visibility into equipment health or degradation patterns.

Impact: 65% of maintenance was unplannedClient-reported baseline

Siloed systems

Over 12 different vendor systems (HVAC, lighting, access control, fire safety, elevators) operated independently with no unified view or coordinated control.

Impact: No single source of truthClient-reported baseline

Tenant experience

Temperature complaints, slow elevators, and inconsistent lighting frustrated tenants. The building struggled to retain premium tenants in a competitive market.

Impact: Premium tenants at risk of leavingClient-reported baseline

Impact figures describe the situation before the project, as the client reported it to us; we did not measure them.

Our solution

We designed the sensing and edge layer of a digital-twin platform — custom sensor units, floor-level gateways and protocol translation — and the data pipeline that keeps the virtual model of the building in sync with its systems.

System architecture

Multi-layer architecture integrating physical IoT sensors with a cloud-based digital twin engine and AI-driven control. The digital twin engine and AI optimisation layers are the client's platform; we built the physical and edge layers and the data pipeline into it.

Physical layer

  • Custom multi-sensor units across building systems
  • BLE beacons for occupancy and asset tracking
  • Smart meters for granular energy monitoring
  • Environmental sensors (CO2, PM2.5, humidity, temperature)
  • Vibration sensors on critical HVAC equipment

Edge computing layer

  • Floor-level edge gateways (NVIDIA Jetson)
  • Real-time data aggregation and preprocessing
  • Local AI inference for time-critical decisions
  • BACnet/Modbus protocol translation
  • Ethernet uplink with 5G backup

Digital twin engine (client platform, not our scope)

  • 3D BIM model integration (Autodesk Platform Services, formerly Forge)
  • Real-time physics simulation (thermal, airflow)
  • What-if scenario modeling
  • Historical pattern analysis
  • Multi-system correlation engine

AI optimization layer (client platform, not our scope)

  • Reinforcement learning for HVAC optimization
  • Predictive maintenance ML models
  • Demand forecasting algorithms
  • Anomaly detection neural networks
  • Natural language interface for facility managers

Custom hardware design

Edge Gateway
NVIDIA Jetson Orin module, one per floor
Sensor Nodes
Custom ESP32-S3 based multi-sensor units
Protocol Support
BACnet, Modbus, KNX, MQTT, OPC-UA
Network
ESP32-S3 units on Wi-Fi/BLE to floor gateways; gateways on Ethernet with 5G backup
Redundancy
Dual-path connectivity, UPS backup
Cybersecurity
Hardware security modules, zero-trust architecture

Edge intelligence features

  • Sub-100ms local decision making for critical systems
  • Federated learning for privacy-preserving model updates
  • Automatic protocol discovery and device onboarding
  • Secure boot and encrypted firmware updates
  • Local buffering for 7 days of offline operation

AI-powered digital twin intelligence

These models belong to the client's digital-twin platform (not our scope) and run on our gateways and in the cloud. The figures below are the platform's design targets.

Thermal comfort optimization

Deep Reinforcement Learning (DQN)

94% occupant comfort prediction

Design target

Continuous optimization, 5-min control intervals

Predictive maintenance

LSTM + Attention mechanism

91% failure prediction accuracy

Design target

Occupancy prediction

Transformer-based time series model

96% accuracy for next-day prediction

Design target

Energy demand forecasting

Gradient Boosting + Weather integration

±3% day-ahead demand prediction

Design target

Anomaly detection

Variational Autoencoder (VAE)

97% anomaly detection rate

Design target

Implementation timeline

Phase durations are approximate and cover the main engineering work only, so they add up to less than the 14 months programme.

Phase 1: Discovery & BIM integration

8 weeks
  • Building systems audit and documentation
  • Mapping sensor points to the client's BIM model
  • Network infrastructure assessment
  • Stakeholder workshops and requirements gathering

Phase 2: IoT infrastructure deployment

12 weeks
  • Sensor network design and installation
  • Edge gateway deployment floor by floor
  • Protocol integration with legacy systems
  • Network security implementation

Phase 3: Data pipeline & dashboards

16 weeks
  • Data pipeline into Azure Digital Twins
  • Edge model deployment on the floor gateways
  • Dashboard and visualization development

Phase 4: Optimization & handover

10 weeks
  • AI model fine-tuning with live data
  • Support for the client's control-system commissioning
  • Staff training and change management
  • Performance baseline establishment

Results & impact

The building owner tracks energy, maintenance and occupant outcomes on its own platform; those results are not published in this case study. Our deliverables were the sensor units, edge gateways, protocol integration and data pipeline described above.

How to read the labels: Design target = an engineering target, not a field measurement; Client-reported baseline = the client's own figure for the situation before the project; we did not measure it. Figures without a label are specifications or project scope.

Technologies used

  • NVIDIA Jetson
  • ESP32-S3
  • Azure Digital Twins
  • TensorFlow
  • PyTorch
  • BACnet
  • Modbus
  • 5G NR
  • Grafana
  • TimescaleDB
  • React
  • Three.js
  • Kubernetes

Related case studies

Manufacturing / Heavy Industry

Industrial IoT: Predictive maintenance sensors and edge AI for a steel plant

Infrastructure / Civil Engineering / Smart Cities

Battery-free IoT: Energy-harvesting sensor nodes for structural health monitoring

Related services, industries and guides