IN-24-C013 - Integrating IR, HDR, and Lidar Cameras into Building Controls PDF

The research team developed and tested a LiDAR and infrared-based thermostat. This thermostat stacks multiple images and uses a neural network to measure the volume of the space, the location of the occupant, and generate a Predicted Mean Vote for occupant comfort based on ASHRAE standard 55. A miniature High Dynamic Range camera was incorporated to analyze lighting conditions to dim lights or operate blinds to minimize glare. The purpose of the device is to use machine learning as a more accurate method of capturing occupancy levels and scale ventilation based on ASHRAE 62.1 or 241 requirements. The point cloud data measures the volume of a space, and when stacked with an infrared image, the location of an occupant is captured using a Convolution Neural Network. The device is only slightly larger than a conventional thermostat and can be placed on a wall. The miniature cameras are mounted on a RaspberryPi, which collects and processes the radiometric and spatial data for thermal comfort parameters using Python as the main operating language. The outputs (heating and cooling signals and occupancy status) are converted to BACnet IP signals that can be passed to a building energy management system. Benchtop testing was conducted in an enclosed office environment to analyze the performance of the Infrared/LiDAR Thermostat, with an occupant seated at the center of the room working at their desk. PMV parameter outputs from the Infrared Thermostat complied with ASHRAE Standard 55-2020 and successfully interacted with unitary HVAC equipment in the office. While Carbon Dioxide monitors are often used for demand-controlled ventilation, the integration of machine learning with LiDAR and Infrared cameras provides enhanced security, less calibration maintenance, and more accurate occupancy levels to efficiently scale ventilation in buildings.