Many drivers are familiar with the irritation of being stuck in traffic on a sweltering summer day. Two researchers at the University of Michigan are working to make uncomfortable situations like this a bit more bearable, by developing a system that will automatically control the climate within a car to optimize both the passengers’ comfort level and the efficiency of the HVAC system.
Over the past few years, Mohamed Abouelenien and Mihai Burzo have been developing approaches to analyze and detect various human behaviors, including lying, feeling stressed, remaining alert at the wheel, and expressing affection, among others. Their latest effort has been to develop a system for cars and homes that automatically detects a person’s thermal discomfort and adjusts accordingly, without any human input.
Abouelenien says there are multiple benefits of such a system that extend beyond creating a comfortable environment for passengers. Notably, raising temperatures by just a few degrees Celsius can result in energy savings and increase the efficiency of HVAC systems, he says. “More importantly, a driver with a thermally comfortable sensation will be less stressed, less fatigued, and more alert, which results in safer driving conditions for the vehicle’s occupants as well as for pedestrians.”
But what temperature yields the best comfort level? At a laboratory at the University of Michigan, the researchers had 50 participants sit in a thermally controlled enclosure while they were exposed to air temperatures ranging between 16 degrees Celsius (61 degrees Fahrenheit) and 35 degrees C (95 degrees F). Participants rated their comfort levels as a remote heat-sensing tool and four types of contact-based physiological sensors collected data describing their heart rate, skin temperature, respiration rate, and skin conductance.
The experimental station includes an insulating enclosure, physiological sensors, and thermal cameras.Photos: Mihai Burzo/IEEE
From the thermal imaging data, the researchers segmented participants’ faces, identified interesting points for tracking, and then contrasted thermal maps of each face. Using this data and a total of 59 physiological features captured by the four contact sensors, they applied machine learning algorithms to automatically detect the thermal sensation of the participants. Then they introduced a cascaded machine learning system that further detected different levels of hot and cold discomfort.
Their results show that thermal imaging was sufficient in detecting the discomfort levels of the study participants—but the efficiency of detection was increased by 18.5 percent when the other physiological features are accounted for.
Drivers, however, are probably not interested in wearing the contact sensors while they commute. Now, Abouelenien and Burzo are working on extracting the physiological factors from the thermal images, which could lead to a fully non-contact detection system. They say several companies have expressed interest in this technology.
This recent work, published in IEEE Intelligent Systems on 30 August, also reveals some interesting insights into temperature comfort. “The time duration needed to reach a certain level of cold discomfort is approximately double that is needed (to reach) the hot sensation, which indicates that human bodies have faster adaptation to heat,” Abouelenien says. He also notes that while he expected the skin temperature sensors to be one of the more reliable indicators of discomfort, in some cases the heart rate features were a more accurate indicator of discomfort.
Besides developing the system to be fully non-contact, the researchers plan to explore other thermal comfort factors such as humidity, clothing level, and metabolic rate. They are also faced with the challenge of adapting this technology so that it accounts for multiple passengers or inhabitants.