A module for TrajSuite. Algorithms to automatically discover stops and moves in GPS trajectories.

White-paper describing POSMIT can be found here.


  • POSMIT - Probability of Stops and Moves in Trajectories. Uses a probabilistic approach to determine if a given entry in the trajectory is stopping or moving. Unlike traditional approaches POSMIT allows the user filter out low probability stops using the minimum stop probability parameter. Also, POSMIT comes with some heuristics to estimate its parameters.
  • CB-SMoT - Clustering-based Stop and Moves of Trajectories. Uses a modified DB-SCAN algorithm to find stops.
  • SMoT - Stops and Moves of Trajectories. Uses predefined regions to search for a minimum duration stay within those region. Unlike the original algorithm our implementation creates these regions by divided up the studying region into uniform cells. The original algorithms calls for the author to pass in interesting regions.


Below are some code usages for each of the algorithms provided in this repo.


//load in a trajectory to process
STTrajectory traj = ....
//how many entries to search either side of each entry 
int nSearchRadius = 2;
//how much a stop can spatially jitter around and still be considered a stop (remember GPS is quite noisy)
double stopVariance = 1.5;
//intialise the actual POSMIT algorithm
POSMIT algo = new POSMIT();
//find the stop probability for each entry in the trajectory
double[] stopProbabilities = algo.run(traj, nSearchRadius, stopVariance);
//set a threshold for an entry to be classified as a stop, otherwise it is a move.
//i.e in this case only entries with an 80% or higher probability become stops,
//whilst lower probability entries become moves.
double minStopConfidence = 0.8;
//convert the trajectory into a trajectory with stop/move annotations at each entry
STStopTrajectory stopTraj = algo.toStopTrajectory(traj, stopProbabilities, minStopConfidence);

*Note: see FindStopsPOSMIT.java for a full example.


//load in a trajectory to process
STTrajectory traj = ....
//Much like DB-SCAN you must specify a spatial epsilon to control cluster growth
double epsMeters = 1.5;
//CB-SMoT introduces a concept that a cluster can only become a stop if it has a 
//total duration equal to or exceeding a user-specified minimum stop duration
long minTimeMillis = 3000L;
//run the CB-SMoT algorithm
STStopTrajectory outTraj = new CBSMoT().run(traj, epsMeters, minTimeMillis);

*Note: see FindStopsCBSmot.java for a full example.

Working with the source

The source is licensed under the MIT licsense so feel free to use it in your projects. It does have some dependencies which are listed in the build.gradle file. The easiest use-case is setting the source up as a gradle project and letting gradle grab those dependencies for you. Next easiest is maven, though you will have translate the dependencies yourself.

Using this project as a library in your own project, your build.gradle file will have to include these:

repositories {
    maven{url 'https://dl.bintray.com/lukehb/137-stopmove'} //hosted on bintray

dependencies {
    compile 'onethreeseven:stopmove:0.0.4'

....Or without cloning, the built source is also hosted on BinTray and can be downloaded: Download

137 Stopmove

Algorithms to automatically discover stops and moves in GPS trajectories.

137 Stopmove Info

⭐ Stars19
🔗 Source Codegithub.com
🕒 Last Updatea year ago
🕒 Created6 years ago
🐞 Open Issues1
➗ Star-Issue Ratio19
😎 Authorlukehb