Crowd Flow Estimation and Prediction using Passive WiFi Data

Project Summary:
Our campus is getting more and more crowded. But can we know how crowded it is now in the library, a particular classroom, or a favorite canteen? In this project, we aim to build a technological solution to crowd flow estimation and prediction by leveraging log data of the campus-wide WiFi network. WiFi networks record client activities, e.g., the time when a client is connected/disconnected from an Access Point (AP). These data allow us to infer the crowd density at different places with advanced data analytics. We explore this opportunity and investigate algorithms to (1) estimate the real-time crowd density at a particular place from the WiFi AP data, (2) analyze the historical crowd flow for different places, and (3) predict the crowd flow variations, which together deliver a campus version of the “Popular times” feature in Google Maps. The project will use privacy-insensitive passive WiFi logs without any cooperation from the users. We will develop the core algorithm based on offline data traces and build a prototype system to demonstrate the performance. We will then explore collaborations with HKU IT support to develop a real-time App/Web to offer “Popular times” as a smart campus service.

Deliverables:
may include a technical report (or an academic paper), a feasibility report, and a prototype system.

Preferred discipline(s):
CS, EEE, or related fields

Project Essential Skills:
strong programming skills, basic knowledge of data analytics and AI, background in algorithms, basic knowledge of wireless networking.

Other Selection Criteria (if any):
motivated

Details of supervision arrangements:
Physically in the semester, while hybrid mode is also acceptable.