• Constraints. Constraints represented as formulas can be used to
provide implicit action preconditions (e.g., (GKL97)). Namely, the set A( s) of actions
applicable in s will exclude all actions a that lead to states Sa = next( a, s) that violate
a constraint. Such con
straints can be used to express capacity constraints (e.g., that the number of balls being
held cannot exceed the number of grippers) or control knowledge (BK98). With constraints, any
Constraint Satisfaction Problem (CSP) can be expressed as a Planning prob lem. The question is
how to make those constraints play an active
role in the search. So far, only approaches based exclusively on constraint techniques are able
to do that (e.g., (KS99; BC99)). A combination of constraint-directed and heuristic search
techniques may also prove useful.