On the feasibility of COVID-19 proximity detection using Bluetooth low energy signals

The Corona VIrus Disease 2019 (COVID-19) pandemic has had a profound effect on the entire world. With the onset of the pandemic in 2020, also started various efforts around the world to automate the contact tracing process to increase its efficacy. Most of these efforts and the smartphone apps that were developed used the Bluetooth Low Energy (BLE) Received Signal Strength Indicator (RSSI) to detect proximity between people. This was facilitated by the development of the Google-Apple Exposure Notification (GAEN) system that made it possible for Android phones and iPhones to exchange standardized BLE messages seamlessly so that RSSI could be measured by both types of phones. In the summer of 2020, we carried out a five-week long BLE RSSI data collection campaign using wearable devices developed at NIST and running on a Raspberry Pi platform. The data collection was comprehensive, because it included a wide variety of operational scenarios including situations where the two devices were not in line-of-site of each other. To the best of our knowledge, there are no publicly reported repositories of BLE RSSI data that includes non-line-of-site (NLOS) scenarios. Such scenarios are important, because they can lead to false alarms in detecting situations where virus transmission between two people is possible. The paper presents a classical analytical framework for proximity detection in the context of electronic contact tracing for COVID-19. We make a distinction between instantaneous and after the fact proximity detection. In the former case, the purpose is to instantaneously warn people that they are violating social distancing rules. In the latter case, the purpose is to notify people after the fact that they had a close contact of certain duration with an individual who has tested positive for COVID-19 and was capable of transmitting the virus at the time of the contact. We evaluate the performance of various methods for instantaneous proximity detection. In after the fact proximity detection, we are interested in identifying periods of time where two people were too close to each other with no barriers between them. We propose a method that applies the Viterbi Algorithm to a time series of BLE RSSI data to solve this problem. The method exploits the fact that humans do not move faster than certain speed. In both cases of instantaneous and after the fact proximity detection, we show that methods based on BLE RSSI data leave a lot of room for improvement. Therefore, the problem of proximity detection for electronic contact tracing and blunting the spread of highly infectious diseases is far from solved. We also know that the next pandemic is not a matter of if but when.