CrowdTracing: Overcrowding Clustering and Detection System for Social Distancing Eiman Kanjo, Dario Ortega Anderez, Amna anwar Smart Sensing Lab Department of Computer Science Nottingham Trent University, Nottingham, UK eiman.kanjo@ntu.ac.uk Ahmad Alshami Department of Computer Science Southern Arkansas University Magnolia, Arkansas, USA James William Department of Computer Science University of Nottingham Nottingham, UK Abstract—Maintaining social distancing in public spaces plays a pivotal role in decreasing COVID-19 contagion and viral spread. COVID-19 has required many countries around the world to close work places, schools and public spaces. This has prompted policy makers, venue managers and local authorities to investigate practical mitigation strategies using technology to exit the lockdown safely and enable the reopening of cities and public spaces. This paper introduces CrowdTracing, a dynamic overcrowding detection system that encourages social-distancing and triggers an alert to venue, city council or facility managers in a dynamic and privacy-preserving manner. CrowdTracing utilises ubiquitous WiFi probing and density-based clustering techniques which can be performed in real-time to identify commonly crowded areas and assist in the estimation of excess gatherings. The proposed system can also be used to enable discovery of where social distancing rules are not being followed, enabling a rapid response, controlling or slowing down the spread of the virus. A classification recall of 0.85 on an experiment with 1000 simulated scenarios were achieved. This indicates the CrowdTracing system proposed was able to identify 85 out 100 scenarios in which social distancing rules were not being followed. Index Terms—COVID-19, Ubiquitous Computing, Wireless probe requests, Mobile Sensing, Clustering, People Count, DB- SCAN, SOM I. I NTRODUCTION The COVID-19 outbreak has changed the way people navigate the world. A large proportion of the population now work or study from home and purchase items over the internet. The situation has the potential to cause significant harm to local economies, affecting high streets, shopping centres, and educational facilities. As local authorities open their venues and city centres and communities begin to return to work and school, the focus has shifted from avoiding contact completely to ensuring social distancing is followed to prevent further outbreaks of COVID-19 [26] [4]. Many policy makers and local authorities are rapidly devel- oping strategies to recover and reopen cities and high streets. In particular they are looking for robust technologies to moni- tor public spaces and disperse overcrowding. For example, the Federal Office of Public Health (FOPH) has requested analyses from Swisscom to check whether the measures to protect against COVID infections are being observed. To make crowds of people visible, the Swisscom Mobility Insight platform Identify applicable funding agency here. If none, delete this. Fig. 1: Illustration of gatherings with and without social distancing. identifies areas of 100 by 100 meters with at least 20 SIM cards, by using approximate location data from the previous day [15]. Conventional crowd monitoring often focuses on increasing footfall and activity public places such as high streets [24], however, due to the recent outbreak, the focus has shifted towards protecting city dwellers by monitoring compliance with social distancing practices and detecting “hotspots” of overcrowding activities. Another project [32] attempted to pre- dict the maximum capacity of a public space when following social distancing and dynamically plotting the distribution of crowds. Thermal and laser ranging sensors are utilised in Singapore to maintain and enforce social distancing in some venues [29]. To tackle the overcrowding issue in city spaces, manual crowd counting and social distancing applications [14] have seen significant usage in an attempt to prevent the spread of COVID-19 through Contact Tracing systems. However, while these systems are effective to alert individuals who have been in contact with a new positive case of COVID-19 in the last days, such alert takes place when the infection may have already happened. Thus, there is a need for an automated social distancing detection approaches to allow for a quicker and more effective response. Traditional footfall counting systems such as cameras suffer from issues related to area coverage, cost, and lack of information about people’s movement. With the recent advancement of short-range wireless technologies such as Bluetooth and WIFI, new data sources have emerged