We present an experimental comparison of three modified DeltaBlue algorithms for local-propagation-based constraint solving. Our three modified methods are respectively called DeltaDown method, DeltaUp method and DeltaCost method. These methods were designed to speed up the planning phase or the evaluation phase of the original DeltaBlue method using additional cost functions to break a tie of the walkabout strength. Our cost functions are respectively called up cost and down cost. These cost functions can give us information about the upstream and the downstream constraints. Our experiments show that DeltaUp method brings us a considerable improvement of the total performance of DeltaBlue method using a small overhead of keeping the cost function.
[1]
Alan Borning,et al.
ThingLab: a constraint-oriented simulation laboratory
,
1979
.
[2]
Bjørn N. Freeman-Benson,et al.
Constraint Hierarchies
,
1992,
LISP Symb. Comput..
[3]
Wm Leler,et al.
Constraint programming languages
,
1987
.
[4]
William Leler,et al.
Constraint programming languages: their specification and generation
,
1988
.
[5]
Bjørn N. Freeman-Benson,et al.
An incremental constraint solver
,
1990,
CACM.
[6]
John H. Maloney.
Using constraints for user interface construction
,
1992
.