Consensus Based Covid-19 Hotspot Network Size Estimation And Node Counting

The primary focus of this work is to detect transmission hotspots where the contact between people infected with COVID-19 and those uninfected is higher than average. We design consensus-based distributed detection and estimation strategies to estimate the size, density, and locations of COVID-19 hotspots to enable well-informed advisories based on data-driven continuous risk assessment. We assume every person has a mobile device and use data such as Bluetooth and WiFi from them to detect transmission hotspots. In order to estimate the number of people in a specific outdoor geographic location and their proximity to each other, we first perform consensus-based distributed clustering to group people into sub-clusters and then estimate the number of users in a cluster. Our algorithm has been configured to be privacy-preserving and operates indoors where we consider signal attenuation and clutter. Here, we demonstrate our GUI that performs network size estimation and node counting over a specific area in real-time, only using local communications. The GUI has controls to perform clustering using the DBSCAN algorithm and then select a cluster and run distributed algorithms to obtain network size and node count. Our results on outdoor hotspot simulations consistently show an accurate estimate of the number of people in a region and their proximity. This data can be used by researchers to study the spread of COVID-19, and by businesses and policymakers to ensure the safety of employees and customers. In future work, we will use data from devices to estimate the center and the radius of each cluster to determine the area covered, the number of devices within the area, and the density of the number of devices, providing complete information of the hotspot. This collaborative work is funded by an NSF RAPID grant and involves two I/UCRCs, CITeR (CU) and SenSIP (ASU).