org.opentripplanner.profile

Class RepeatedRaptorProfileRouter

java.lang.Object

org.opentripplanner.profile.RepeatedRaptorProfileRouter

public class RepeatedRaptorProfileRouter
extends Object

Perform one-to-many profile routing using repeated RAPTOR searches. In this context, profile routing means finding the optimal itinerary for each departure moment in a given window, without trying to reconstruct the exact paths. In other contexts (Modeify-style point to point routing) we would want to include suboptimal but resonable paths and retain enough information to reconstruct all those paths accounting for common trunk frequencies and stop clusters. This method is conceptually very similar to the work of the Minnesota Accessibility Observatory (http://www.its.umn.edu/Publications/ResearchReports/pdfdownloadl.pl?id=2504) They run repeated searches for each departure time in the window. We take advantage of the fact that the street network is static (for the most part, assuming time-dependent turn restrictions and traffic are consistent across the time window) and only run a fast transit search for each minute in the window.

Field Summary

Fields

Modifier and Type	Field and Description
String	banAgency If not null, completely skip this agency during the calculations.
Graph	graph
boolean	includeTimes
static int	MAX_DURATION
RaptorWorkerData	raptorWorkerData If this is null we will generate a throw-away raptor data table.
ProfileRequest	request
int	stepMinutes
TimeSurface.RangeSet	timeSurfaceRangeSet Three time surfaces for min, max, and average travel time over the given time window.
TaskStatistics	ts

Constructor Summary

Constructors

Constructor and Description
RepeatedRaptorProfileRouter(Graph graph, ProfileRequest request) Deprecated.
RepeatedRaptorProfileRouter(Graph graph, ProfileRequest request, SampleSet sampleSet) Make a router to use for making ResultSets.

Method Summary

Modifier and Type	Method and Description
gnu.trove.map.TIntIntMap	findInitialStops(boolean dest, RaptorWorkerData data) Find all transit stops accessible by streets around the origin, leaving behind a shortest path tree of the reachable area in the field preTransitSpt.
static RaptorWorkerData	getRaptorWorkerData(ProfileRequest request, Graph graph, SampleSet sampleSet, TaskStatistics ts) Create RAPTOR worker data from a graph, profile request and sample set (the last of which may be null
ResultEnvelope	route()

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

MAX_DURATION

public static final int MAX_DURATION

See Also:

Constant Field Values

request

public ProfileRequest request

graph

public Graph graph

stepMinutes

public int stepMinutes

timeSurfaceRangeSet

public TimeSurface.RangeSet timeSurfaceRangeSet

Three time surfaces for min, max, and average travel time over the given time window.

banAgency

public String banAgency

If not null, completely skip this agency during the calculations.

raptorWorkerData

public RaptorWorkerData raptorWorkerData

If this is null we will generate a throw-away raptor data table. If it is set, the provided table will be used for routing. Ideally we'd handle the cacheing of reusable raptor tables inside this class, but this class is a throw-away calculator instance and doesn't have access to the job IDs which would be the cache keys.

ts

public TaskStatistics ts

includeTimes

public boolean includeTimes

Constructor Detail

RepeatedRaptorProfileRouter

@Deprecated
public RepeatedRaptorProfileRouter(Graph graph,
                                               ProfileRequest request)

Deprecated.

Make a router to use for making time surfaces only. If you're building ResultSets, you should use the below method that uses a SampleSet; otherwise you maximum and average may not be correct. If you want isochrones, you should use this method, or use a sampleset that is very fine. The reason for this is that isochrones are interpolated from these points in Euclidean space, so the points need to be very fine. Consider the case of three roads in an equilateral triangle, the edges of which each take ten minutes to traverse. If you're standing at one vertex and want to go halfway down the opposite edge, it is a fifteen minute trip. However, interpolating between the vertices will not give fifteen minutes, it will give ten. The less granular your representation is, the worse this problem will be. TODO merge this with the 3-arg constructor, and just specify what happens when you have a null pointset. This should allow us to get rid of some conditionals in the calling code.

RepeatedRaptorProfileRouter

public RepeatedRaptorProfileRouter(Graph graph,
                                   ProfileRequest request,
                                   SampleSet sampleSet)

Make a router to use for making ResultSets. This propagates times all the way to the samples, so that average and maximum travel time are correct. Samples are linked to two vertices at the ends of an edge, and it is possible that the average for the sample (as well as the max) is lower than the average or the max at either of the vertices, because it may be that every time the max at one vertex is occurring, a lower value is occurring at the other. This initially seems improbable, but consider the case when there are two parallel transit services running out of phase. It may be that some of the time it makes sense to go out of your house and turn left, and sometimes it makes sense to turn right, depending on which is coming first.

Method Detail

route

public ResultEnvelope route()

findInitialStops

public gnu.trove.map.TIntIntMap findInitialStops(boolean dest,
                                                 RaptorWorkerData data)

Find all transit stops accessible by streets around the origin, leaving behind a shortest path tree of the reachable area in the field preTransitSpt.

Parameters:

data - the raptor data table to use. If this is null (i.e. there is no transit) range is extended, and we don't care if we actually find any stops, we just want the tree of on-street distances.

getRaptorWorkerData

public static RaptorWorkerData getRaptorWorkerData(ProfileRequest request,
                                                   Graph graph,
                                                   SampleSet sampleSet,
                                                   TaskStatistics ts)

Create RAPTOR worker data from a graph, profile request and sample set (the last of which may be null