Discovering Real-time Reachable Area using Trajectory Connectioins
Finding the Reachable Area of a Specified Location within a Given Time Period in Real-time Conditions.
Demo System | Experimental Code


Key Features

(1) Indexing Meets Machine Learning

We give a good example for combining indexing with machine learning techniques to find real-time reachable areas.

(2) High Efficiency

We design a set of indexing and query processing techniques to prune redundant trajectory connections, which can efficiently find real-time reachable areas with arbitrary connection numbers.

(3) Intelligent Connection Number Selection

We train a spatio-temporal model to predict the best trajectory connection number in any location and at any time.


Data Settings

The following picture displays the spatio-temporal distribution of trajectories in Shanghai. Figure (a) indicates that downtown areas contain more taxi activities than suburb areas. Figure (b) presents that there are more trajectories during rush hours. To this end, a dynamic connection number needs to be predicted when a reachable area query arrives.
* Spatio-temporal Distribution of Trajectories in Shanghai, China.


Our method shows effectiveness and efficiency for the real-time reachable area discovery.

Experimental Code

Experimental Code

System Demostration

User Interface

Users can access the demo system through: The user interface is shown as the following picture, which has three panels: ① Selection Panel, ② Button Panel, and ③ Map Panel

Operation Demo


Pseudocode of TE

Pseudocode of SG-index Construction

Correlations Between k-value Distribution and External Factors

Figure (a) shows that with a smaller traffic flow, less connections should be considered.
Figure (b) indicates that comparing to sunny, more trajectory connections need to be considered when it goes rainy.
Figure (c) illustrates that a denser POI distribution results in more trajectory connections.
Figure (d) shows more intersections lead to a bigger k value.