org.opentripplanner.profile

Class PropagatedTimesStore

java.lang.Object

org.opentripplanner.profile.PropagatedTimesStore

public class PropagatedTimesStore
extends Object

Stores travel times propagated to the search targets (all vertices in the street network or a set of points of interest) in one-to-many repeated raptor profile routing. Each raptor call finds minimum travel times to each transit stop. Those must be propagated out to the final targets, giving minimum travel times to each street vertex (or each destination opportunity in accessibility analysis). Those results are then merged into the summary statistics per target over the whole time window (over many RAPTOR calls). This class handles the storage and merging of those summary statistics. Currently this includes minimum, maximum, and average earliest-arrival travel time for each target. We could conceivably retain all the propagated times for every departure minute instead of collapsing them down into three numbers. If they were all sorted, we could read off any quantile, including the median travel time. Leaving them in departure time order would also be interesting, since you could then see visually how the travel time varies as a function of departure time. We could also conceivably store travel time histograms per destination, but this entails a loss of information due to binning into minutes. These binned times could not be used to apply a smooth sigmoid cutoff which we usually do at one-second resolution. When exploring single-point (one-to-many) query results it would be great to have all these stored or produced on demand for visualization.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	PropagatedTimesStore.ConfidenceCalculationMethod

Field Summary

Fields

Modifier and Type	Field and Description
int	N_BOOTSTRAPS

Constructor Summary

Constructors

Constructor and Description
PropagatedTimesStore(Graph graph, ProfileRequest req)
PropagatedTimesStore(Graph graph, ProfileRequest req, int size)

Method Summary

Modifier and Type	Method and Description
int	countTargetsReached()
ResultEnvelope	makeIsochronesForVertices() This bypasses a bunch of conversion and copy steps and just makes the isochrones.
ResultEnvelope	makeResults(SampleSet ss, boolean includeTimes, boolean includeHistograms, boolean includeIsochrones) Make a ResultEnvelope directly from a given SampleSet.
SparseMatrixZSampleGrid<SampleGridRenderer.WTWD>	makeSampleGridForVertices(int[] times, double gridSizeMeters) Create a SampleGrid from only the times stored in this PropagatedTimesStore.
TimeSurface.RangeSet	makeSurfaces(RepeatedRaptorProfileRouter repeatedRaptorProfileRouter)
void	setFromArray(int[][] times, boolean[] includeInAverages, PropagatedTimesStore.ConfidenceCalculationMethod confidenceCalculationMethod)

Methods inherited from class java.lang.Object

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

Field Detail

N_BOOTSTRAPS

public final int N_BOOTSTRAPS

See Also:

Constant Field Values

Constructor Detail

PropagatedTimesStore

public PropagatedTimesStore(Graph graph,
                            ProfileRequest req)

PropagatedTimesStore

public PropagatedTimesStore(Graph graph,
                            ProfileRequest req,
                            int size)

Method Detail

setFromArray

public void setFromArray(int[][] times,
                         boolean[] includeInAverages,
                         PropagatedTimesStore.ConfidenceCalculationMethod confidenceCalculationMethod)

Parameters:

times - for search (varying departure time), an array of travel times to each destination.

includeInAverages - for each iteration, whether that iteration should be included in average calculations. In RaptorWorker's Monte Carlo code we also include minima and maxima, which should not be included in averages. Iterations that are not included in averages are still used to determine extrema.

makeResults

public ResultEnvelope makeResults(SampleSet ss,
                                  boolean includeTimes,
                                  boolean includeHistograms,
                                  boolean includeIsochrones)

Make a ResultEnvelope directly from a given SampleSet. The RaptorWorkerData must have been constructed from the same SampleSet.

makeSurfaces

public TimeSurface.RangeSet makeSurfaces(RepeatedRaptorProfileRouter repeatedRaptorProfileRouter)

makeIsochronesForVertices

public ResultEnvelope makeIsochronesForVertices()

This bypasses a bunch of conversion and copy steps and just makes the isochrones. This assumes that the target indexes in this router/propagatedTimesStore are vertex indexes, not pointset indexes. TODO parameter for a pointset or a vertex lookup table, so we can handle both.

makeSampleGridForVertices

public SparseMatrixZSampleGrid<SampleGridRenderer.WTWD> makeSampleGridForVertices(int[] times,
                                                                                  double gridSizeMeters)

Create a SampleGrid from only the times stored in this PropagatedTimesStore. This assumes that the target indexes in this router/propagatedTimesStore are vertex indexes, not pointset indexes. This is not really ideal since it includes only intersection nodes, and no points along the road segments. FIXME this may be why we're getting hole-punching failures. TODO: rewrite the isoline code to use only primitive collections and operate on a scalar field.

countTargetsReached

public int countTargetsReached()