CROMO: Enhancing Crowd Mobility Characterization through Real-time Radio Frequency Data Analytics

Abstract

Human casualties at entertaining, religious, or political events often occur due to lack of proper crowd management. Notably, for the crowd in mobile, a minor accident can create a panic for the people to start stampeding and trampling others. Although many smart video surveillance technologies are recently proposed, it is still very challenging problems to predict a crash in real-time among the mobile crowd for preventing any potential disaster.In this paper, we propose CROMO that enhances crowd mobility characterization through real-time Radio Frequency (RF) data analytics. Inspired by the recent advanced artificial intelligence (AI) technology and machine learning (ML) algorithms, traditional video surveillance technologies make object detection and identification possible in real-time. However, their scalability and capacity lack in a crowded mobile environment. CROMO propose to fill the gap via RF signal analytics. Among the many crowd mobility characteristics, we tackle object group identification, the speed, and direction detection for the mobile group. We also apply them to group semantics to track the crowd status and predict any potential accidents and disasters. Taking advantage of power-efficiency, cost-effectiveness, and ubiquitous availability, we specifically analyze a Bluetooth Low Energy (BLE) signal. We have tested CROMO in both a practical crowd event and the controlled indoor and outdoor lab environments. The results show that CROMO can detect the direction, the speed, and the density of the mobile crowd in real-time. Therefore, it can help the crowd management in avoiding disasters possibilities at crowd events.