How VAUD Lets You Find Lost Phones in Taxis Without Writing a Single Query

Imagine taking a taxi from a train station, losing your phone in the vehicle, and then being able to locate it without writing any code. VAUD (Visual Analysis for Urban Data) makes this possible by turning city‑scale spatio‑temporal queries into an intuitive drag‑and‑drop workflow.

Using taxi and mobile trajectory data, a user inputs the boarding and alighting time and location; the system matches these spatio‑temporal points to taxi trajectories (adding intermediate points improves accuracy) and then links the matched taxi to the phone’s trajectory, revealing whether the phone remains in the vehicle or has been taken by a subsequent passenger. Similar techniques can support police investigations, suspect tracking, and other public‑safety scenarios.

VAUD was built jointly by Zhejiang University’s CAD&CG Lab and Alibaba Cloud’s City Brain team. It combines a data‑analysis module with a query module to provide both data retrieval and inference capabilities. The system first constructs a quad‑tree index and a spatio‑temporal cube model for human and vehicle trajectories, accelerating matching of points that share the same time‑space coordinates.

The query model supports atomic queries based on time (point or interval), space (point or region), or their combinations (intersection, union, difference). Queries can be single‑source spatio‑temporal searches, trajectory‑matching searches, or cross‑domain searches that decompose a complex task into a series of atomic queries and information‑extraction steps.

Visual Query Interface

Through a visual interface, users build query nodes by dragging, dropping, and configuring time‑space conditions, replacing traditional SQL statements. The resulting query chain guides the inference process, lowering the learning curve for non‑technical users. Results are displayed on a map‑based environment‑perception panel, showing detailed attributes such as speed and heat‑map visualizations, enabling interactive investigation.

The system also integrates automated analysis techniques, e.g., using deep‑learning and K‑Nearest‑Neighbor algorithms to classify city base‑station functions and applying dimensionality‑reduction for exploratory interaction.

Potential applications include traffic‑congestion analysis, tide‑road condition monitoring, and public‑safety crowd‑gathering or evacuation planning, helping cities allocate resources more efficiently and maintain healthy urban operations.

Note 1: The experimental data in the paper are from the laboratory, not from Alibaba’s internal or City Brain datasets. The paper “VAUD: A Visual Analysis Approach for Exploring Spatio‑Temporal Urban Data” will appear in IEEE Transactions on Visualization and Computer Graphics, 2018, authored by Huang Zhaosong, Wu Feiran, Zhu Minfeng, and Guan Huihua.

Note 2: Visual analytics combines data mining, computer graphics, perception theory, and human‑computer interaction to accurately convey information and transform data into human insight. The Zhejiang University CAD&CG National Key Lab’s visual‑analysis group, led by Prof. Chen Wei, has long focused on visual data analysis for urban scenarios, and the joint center leverages this expertise together with Alibaba Cloud’s resources.