March 12, 2026 – Cambridge, MA – MIT Pushes PIR Capabilities Beyond Line-of-Sight
Researchers at the Massachusetts Institute of Technology have demonstrated a PIR sensor array capable of detecting occupancy through standard drywall and light masonry walls. The system uses an array of 16 PIR sensors with custom optics and machine learning algorithms to detect the thermal signature of people in adjacent rooms.
The research, published in the Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, represents a significant advancement in non-invasive occupancy sensing.
The Technology
The system, called “ThermoSense,” uses a 4×4 array of high-sensitivity PIR sensors arranged in a 10cm × 10cm panel. Each sensor is equipped with a custom Fresnel lens that creates narrow detection zones, enabling spatial resolution of the thermal environment on the opposite side of a wall.
Key technical elements:
- Sensor array: 16 Panasonic EKMB series PIR sensors
- Optics: Custom 3D-printed Fresnel lenses for each sensor
- Sampling rate: 20 Hz per sensor
- Wall types tested: Drywall (1.5cm), wood (5cm), light masonry (10cm)
- Machine learning: Convolutional neural network trained on 10,000 hours of data
How It Works
The array does not “see through” walls in the traditional sense. Instead, it detects minute temperature variations on the wall surface caused by a person on the other side. A warm body on the opposite side of a wall creates a small temperature gradient on the wall surface (0.1-0.5°C). The array of PIR sensors detects these temperature differences, and the machine learning model interprets them to determine occupancy.
The system cannot create an image, but it can determine:
- Whether a room is occupied
- The approximate number of people (1-3)
- The general location (left/right side of the room)
- Whether occupants are moving or stationary
Performance Results
In controlled tests, the system achieved:
- Occupancy detection accuracy: 94% for single occupant, 87% for two occupants
- Movement detection: 96% accuracy
- Through drywall: Up to 3 meters detection range
- Through wood: Up to 2 meters
- Through light masonry: Up to 1 meter
The system did not work through concrete walls thicker than 15cm, metal studs, or walls with foil-faced insulation.
Applications
Potential applications include:
- Elderly care: Monitor occupancy in adjacent rooms without cameras
- Smart homes: Room-level occupancy for HVAC and lighting without sensors in every room
- Energy management: Detect which rooms are occupied for zoning
- Security: Detect intruders before they enter a room
- Retail analytics: Monitor occupancy in dressing rooms and fitting areas
Limitations and Challenges
The researchers acknowledge several limitations:
- Wall construction: Performance varies significantly with wall material
- Temperature variations: HVAC cycling and sunlight can affect readings
- Cost: 16 sensors plus processing is more expensive than single sensors
- Installation: Requires access to wall surface for sensor placement
- Privacy implications: Detecting through walls raises ethical questions
Research Team Comments
“We’re not trying to replace traditional PIR sensors,” said lead researcher Dr. Emily Chen. “This technology addresses a different use case: detecting occupancy in adjacent spaces without putting sensors in every room. For applications like elderly care, where cameras are intrusive and putting sensors in every room is impractical, this approach offers a middle ground.”
Commercialization Timeline
The research team has filed patents and is exploring commercialization options. Potential pathways include:
- Licensing to existing sensor manufacturers: 2-3 years to market
- Startup company formation: 3-5 years to market
- Integration into building automation systems: 4-6 years
Conclusion
MIT’s through-wall PIR array demonstrates that the capabilities of passive infrared sensing are far from fully explored. While not a replacement for traditional PIR sensors, this technology could enable new applications in privacy-sensitive occupancy monitoring.