David Labrique1, an independent researcher, has a new paper on the Effects of Obstructions on the Accuracy of Bluetooth Contact Tracing.
He looks into the variation of Bluetooth received signal strength (RSSI) due to different types of obstruction such as walls and the human body. He explains how RSSI is being used in contact tracing apps and asks whether it’s possible to have false positives when there’s wall between smartphones or false negatives when people are close together but blocked by their bodies.
David used a Raspberry Pi as a Bluetooth emitter and a smartphone as a receiver, situated 1 metre apart and placed various obstacles between them. He found that drywall and stud walls were ineffective at reducing Bluetooth signal strength. Conversely, human bodies drastically reduce Bluetooth signals.
Smart watches might be possible candidates for more accurate contact tracing as they are less obstructed by the body when worn on the wrist
What David didn’t do was test at different levels of power output. We assume he just used full power which will go through walls. Solutions such as our CATT use lower power predominantly to save battery life but also because there’s no need to transmit further. There’s are also factors at play in a smartphone app such as the variance of signal power across transmitting smartphone, the variance in the ability of different smartphones to receive the signal and the ability (or not) of smartphones to be able to transmit and receive in background when the app isn’t showing/running. These factors make app based contact tracing even more unreliable. Stand alone devices, such as smart watches mentioned by David, work better.