Reasoner
Class ConstraintPropagator

java.lang.Object
  |
  +--Reasoner.ConstraintPropagator

public class ConstraintPropagator

extends java.lang.Object

Performs the position and attitude reasoning over the constraints in a set of object inferences. It breaks the object inferences into disjoint partitions with no interaction and proceeds to solve them independently. Partitions are processed from most to least cardinality to solve the harder, most constraining constraints earlier. Backtracking enforces a systematic approach to dealing with paths that fail to solve all constraints.

The system-level selective non-interpenetration constraint is enforced to prevent the objects in solutions between otherwise independent partitions from occupying the same space.

Version:

Created 3 October 2003

Author:

Dan Tappan

See Also:

SimulationEngine

Nested Class Summary

private class	ConstraintPropagator.Adjacency Maintains an adjacency link from a source object inference to a target object inference with respect to a constraint node.
private class	ConstraintPropagator.Partition Maintains the contents of a partition of object inferences that reference each other and no object inferences in other partitions.

Field Summary

static java.lang.String	ARGUMENT_SELF_PIPPER
static java.lang.String	FIELD_ANYWHERE
static java.lang.String	FIELD_INTERIOR
private MonteCarloRunSet	m_monte_carlo_run_set
private java.util.ArrayList	m_partitions
private java.lang.String	RELATION_IN

Constructor Summary

ConstraintPropagator(MonteCarloRunSet monte_carlo_run_set)
Creates a new constraint propagator to find a solution set from a Monte Carlo run set.

Method Summary

private void	buildPartition(ConstraintPropagator.Adjacency adjacency, java.util.ArrayList adjacencies, ConstraintPropagator.Partition partition) Recursively builds a partition with all interconnected adjacencies.
private void	buildPartitions(MonteCarloRunSet monte_carlo_run_set) Partitions the object inferences into subsets that can be solved independently.
boolean	propagate(int max_partition_iterations, int max_partition_backtracks) Propagates the constraints in each partition.
private void	sortPartitions() Sorts the partitions from largest to smallest.
java.lang.String	toString() Returns a string representation of this class.

Methods inherited from class java.lang.Object

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

Field Detail

FIELD_ANYWHERE

public static final java.lang.String FIELD_ANYWHERE

See Also:

Constant Field Values

FIELD_INTERIOR

public static final java.lang.String FIELD_INTERIOR

See Also:

Constant Field Values

ARGUMENT_SELF_PIPPER

public static final java.lang.String ARGUMENT_SELF_PIPPER

See Also:

Constant Field Values

RELATION_IN

private final java.lang.String RELATION_IN

See Also:

Constant Field Values

m_partitions

private java.util.ArrayList m_partitions

m_monte_carlo_run_set

private MonteCarloRunSet m_monte_carlo_run_set

Constructor Detail

ConstraintPropagator

public ConstraintPropagator(MonteCarloRunSet monte_carlo_run_set)
                     throws java.lang.Exception

Creates a new constraint propagator to find a solution set from a Monte Carlo run set.

Parameters:

monte_carlo_run_set - - the Monte Carlo run set containing the object-inference constraints to solve

Throws:

java.lang.Exception - - if propagation fails

Method Detail

buildPartition

private void buildPartition(ConstraintPropagator.Adjacency adjacency,
                            java.util.ArrayList adjacencies,
                            ConstraintPropagator.Partition partition)

Recursively builds a partition with all interconnected adjacencies.

Parameters:

adjacency - - the adjacency to follow

adjacencies - - the list of adjacencies followed from adjacency

partition - - the partition to add to

See Also:

buildPartitions(Reasoner.MonteCarloRunSet)

buildPartitions

private void buildPartitions(MonteCarloRunSet monte_carlo_run_set)
                      throws java.lang.Exception

Partitions the object inferences into subsets that can be solved independently.

Parameters:

monte_carlo_run_set - - the Monte Carlo run set containing the object-inference constraints to solve

Throws:

java.lang.Exception - - if an object inference referenced by another is not present in the run set

propagate

public boolean propagate(int max_partition_iterations,
                         int max_partition_backtracks)
                  throws java.lang.Exception

Propagates the constraints in each partition.

Returns:

whether propagation succeeded

Throws:

java.lang.Exception - - if propagation fails

sortPartitions

private void sortPartitions()

Sorts the partitions from largest to smallest. This sets the constraint-satisfaction sequence so the partitions that have more constraints, and are therefore more difficult to solve, are considered earlier. This also sorts the contents of each partition to group them.

toString

public java.lang.String toString()

Returns a string representation of this class.

Overrides:

toString in class java.lang.Object

Returns:

the String representation