We are investigating constraint directed heuristic search as a means for performing design in the field of space planning. Space planning is selecting, dimensioning, locating and shaping design units to create two dimensional layouts based on functional, topological and geometrical considerations. Search is carried out using operators at different abstraction levels and design objects at different levels of detail. Constraints are used to represent domain knowledge, to define the search space by specifying operators in means ends analysis manner, and to rate the partial candidate solutions using importances associated with each constraint.

Search is carried out opportunistically. The philosophy behind opportunism is that understanding the approximate topology of the search space will lead to efficient search. Uncertainty associated with constraints is derived and used to identify islands of certainty in the search space, which are used as starting points and anchors for search. The knowledge that enables us to identify opportunistic decisions are interactions between constraints and the usefulness of a constraint in different situations. The resulting uncertainty measure will be tested by observing the problem solving behavior it causes in different search spaces.