Co-Trajectory Analysis

This is a package for performing analysis on collection of trajectories, co-trajectories.

Quick Start

Make sure that the following is installed

• sbt
• scala
• spark

From the project root build it with

sbt package

The first time this is done it downloads a lot of packages and can thus take quite some time. You can perform the, very few, tests with

sbt test

Most of the tests are in pure Scala and should most likely work. Some of the tests are using Spark and if these do not work it could be a problem with you local Spark installation.

To get it running in a local spark-shell use the command

spark-shell --jars target/scala-2.11/co-traj_2.11-1.0.jar --packages com.graphhopper:map-matching:0.7.0,io.spray:spray-json_2.11:1.3.5

You should then be able to run the first example

scala> Examples.example1
Number of trajectories: 5
Number of measurements: 60
Number of possible swaps: 6
Number of possible paths in the DAG: 49
Precomputing data
Computing number of paths for the whole graph
Computing chain of vertices
Computing number of paths trough measurements
Output data about number of paths through measurements to output/example1-1.csv
Number of paths trough the measurement m = MeasurementID(2,Measurement(13,[2.5])): 12
Precomputing data
Computing number of paths between start and end vertex
Number of steps to compute: 5
Every dot is 20 steps, 50 per line
Output data about number of paths starting and ending at a the same vertices as trajectory i to output/example1-2.csv
id = 0: 2
id = 1: 3
id = 2: 2
id = 3: 2
id = 4: 1

The other examples will not run since they need access to external datasets, more information about how to get these can be found further down.

File structure

It is structured as a typical Scala project with the general code in src/main/scala/ and the test code in src/test/scala/. It follows the Scala style of files starting with a capital letter corresponding to a single class, or object, with the same name and files starting with a lower-case letter corresponding to multiple, related, classes or objects. The data/ directory contains data for some co-trajectories used in tests and examples. Output to files is in general done to the output/ directory.

Class structure

The project consists of Classes, which hold data and methods working on that data, and Objects, which only holds methods.

Classes

All the classes are in project are either case classes, essentially tuples with methods on, or implicit classes, renaming of other types with added methods.

The case classes are

• Location(x: Array[Double])
• LocationPartition(x: Array[Int])
• Measurement(time: Long, location: Location)
• MeasurementID(id: Int, measurement: Measurement)
• MeasurementPartition(time: Long, location: LocationPartition)
• MeasurementPartitionID(id: Int, partition: MeasurementPartition)
• Trajectory(id: Int, measurements: Array[Measurement])
• TrajectoryPartition(id: Int, partitions: Array[MeasurementPartition])
• Swap(time: Long, ids: Array[Int])
• LonLatJson(type: String = "MultiPoint", coordinates: Array[(Double, Double)])

The implicit classes are

• CoTrajectory(cotraj: Dataset[Trajectory])
• CoTrajectoryPartition(cotraj: Dataset[TrajectoryPartition])
• Swaps(swaps: Dataset[Swap])

Objects

The objects contain general methods, some of the are helper objects mainly meant for implementing internal methods whereas other for use by the end user. The objects mainly meant for internal use are

• CoTrajectoryUtils
• GraphHopperHelper
• TrajectoryHelper
• SparkSessionHolder

And the ones for general use are

• Parse
• Swapmob
• Examples
• Visualize