Scheduling variable and value ordering heuristics should be improved.
As suggested in Philippe Laborie, Malik Ghallab "Planning with Sharable Resource Constraints", IJCAI 1995: 1643-1651, it is possible to use an heuristic estimator K(φ) such that
K(φ)^-1) = Σ (1 + commit(ρ) - commit(ρmin))^-1
where commit(ρ) represents an estimation of commitment and ranges from 0 to 1. In other words, it estimates the loss in flexibility as a result of posting ρ. ρmin is the constraint with the minimum value of commit(ρ).
Roughly speaking, commit(ρ) is equal to 0 if ρ is redundant with the constraints already posted and is equal to 1 if it introduces an inconsistency.
Intuitively, higher values of K(φ) are preferrable for choosing flaws while smaller values of commit(ρ) are preferrable for choosing resolvers.
Scheduling variable and value ordering heuristics should be improved.
As suggested in Philippe Laborie, Malik Ghallab "Planning with Sharable Resource Constraints", IJCAI 1995: 1643-1651, it is possible to use an heuristic estimator
K(φ)
such thatK(φ)^-1) = Σ (1 + commit(ρ) - commit(ρmin))^-1
wherecommit(ρ)
represents an estimation of commitment and ranges from 0 to 1. In other words, it estimates the loss in flexibility as a result of postingρ
.ρmin
is the constraint with the minimum value ofcommit(ρ)
.Roughly speaking,
commit(ρ)
is equal to0
ifρ
is redundant with the constraints already posted and is equal to1
if it introduces an inconsistency.Intuitively, higher values of
K(φ)
are preferrable for choosing flaws while smaller values ofcommit(ρ)
are preferrable for choosing resolvers.