Package com.graphhopper.jsprit.core.algorithm.recreate

Class AdaptiveSpatialFilter

java.lang.Object

com.graphhopper.jsprit.core.algorithm.recreate.AdaptiveSpatialFilter

public class AdaptiveSpatialFilter
extends Object

Adaptive spatial filter that learns whether spatial filtering is effective for a given problem.

During the learning phase, it compares filtered vs full search results to determine: - Hit rate: How often filtering finds the same best route as full search - Cost deviation: When filtering misses, how much worse is the cost - Miss rate: How often filtering misses feasible insertions entirely

Based on learning results, it adaptively enables/disables filtering: - If filtering works well (high hit rate, low cost deviation): use 95% filtering - If filtering works partially: use 50% filtering with larger k - If filtering doesn't work: disable entirely

Thread-safety: This class is NOT thread-safe. Each thread should have its own instance, or external synchronization is required.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	AdaptiveSpatialFilter.FilterResult	Result of filtering decision.

Constructor Summary

Constructors

Constructor	Description
AdaptiveSpatialFilter()	Creates an adaptive spatial filter with default settings.
AdaptiveSpatialFilter(int learningRounds, int initialK)	Creates an adaptive spatial filter.

Method Summary

Modifier and Type	Method	Description
int	getK()	Returns the current k value (number of nearest routes to consider).
List<VehicleRoute>	getNearestRoutes(Job job, Collection<VehicleRoute> routes, int limit)	Gets the k nearest routes to a job based on minimum distance to any activity.
AdaptiveSpatialFilter.FilterResult	getRelevantRoutes(Job job, Collection<VehicleRoute> routes)	Decides whether to use filtering for this job and returns relevant routes.
String	getStats()	Returns learning statistics for debugging.
boolean	isFilteringEnabled()	Returns whether filtering is currently enabled.
boolean	isLearningComplete()	Returns whether learning is complete.
void	recordComparison(InsertionData filteredBest, VehicleRoute filteredRoute, InsertionData fullBest, VehicleRoute fullRoute)	Records comparison data for learning.
void	reset()	Resets the filter to initial state for a new problem.

Methods inherited from class java.lang.Object

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

Constructor Details

AdaptiveSpatialFilter

public AdaptiveSpatialFilter()

Creates an adaptive spatial filter with default settings.

AdaptiveSpatialFilter

public AdaptiveSpatialFilter(int learningRounds,
 int initialK)

Creates an adaptive spatial filter.

Parameters:

learningRounds - number of comparison rounds for learning

initialK - initial number of nearest routes to consider

Method Details

getRelevantRoutes

public AdaptiveSpatialFilter.FilterResult getRelevantRoutes(Job job,
 Collection<VehicleRoute> routes)

Decides whether to use filtering for this job and returns relevant routes.

During learning phase: returns full routes but marks for comparison. After learning: deterministically applies filtering based on learned effectiveness.

Parameters:

job - the job to find routes for

routes - all available routes

Returns:

FilterResult containing routes to consider and whether comparison is needed

getNearestRoutes

public List<VehicleRoute> getNearestRoutes(Job job,
 Collection<VehicleRoute> routes,
 int limit)

Gets the k nearest routes to a job based on minimum distance to any activity.

recordComparison

public void recordComparison(InsertionData filteredBest,
 VehicleRoute filteredRoute,
 InsertionData fullBest,
 VehicleRoute fullRoute)

Records comparison data for learning. This method is thread-safe for use in concurrent execution.

Parameters:

filteredBest - best insertion from filtered routes (null if none found)

filteredRoute - route for filtered best (null if none found)

fullBest - best insertion from full search (null if none found)

fullRoute - route for full best (null if none found)

isLearningComplete

public boolean isLearningComplete()

Returns whether learning is complete.

isFilteringEnabled

public boolean isFilteringEnabled()

Returns whether filtering is currently enabled.

getK

public int getK()

Returns the current k value (number of nearest routes to consider).

getStats

public String getStats()

Returns learning statistics for debugging.

reset

public void reset()

Resets the filter to initial state for a new problem.