Foundations of Constraint Processing, Fall 2005 November 2, 2005Weak Commitment1 Foundations of Constraint Processing CSCE421/821, Fall 2005: Berthe Y. Choueiry (Shu-we-ri) Avery Hall, Room 123B Tel: +1(402) Note on Least Commitment
Foundations of Constraint Processing, Fall 2005 November 2, 2005Weak Commitment2 Background Constraint propagation –Reduces problem size –Eliminates inconsistent choices –Gets the problem closer to being solved –But does not eliminate any solutions –When it solves the problem, we keep all solutions Search –Solves the problem by making decisions and eliminating perfectly acceptable choices, while keeping one
Foundations of Constraint Processing, Fall 2005 November 2, 2005Weak Commitment3 Generalizing.. Constraint propagation makes no commitment at all Search makes strong commitments Question: anything in the middle? Answer: least commitment –Common: planning, scheduling communities
Foundations of Constraint Processing, Fall 2005 November 2, 2005Weak Commitment4 Least commitment: rationale As long as you do not need to make a commitment, keep propagating If propagation is ‘stuck’ (no progress), –then make the least commitment you need to make in order to enable more propagation without ruling out too many solutions Technique: –typically by adding a constraint (that is weaker than a variable assignment)
Foundations of Constraint Processing, Fall 2005 November 2, 2005Weak Commitment5 Example: scheduling Context: –scheduling tasks in time. You notice the 2 tasks must use the same resource Propagation –adds a mutex constraint between both tasks without ordering them Search –fixes the time for one or both tasks thus ruling out so many possibilities Least commitment –adds a constraint committing one task to be before the other –tasks remain floating in time, –and any task can be inserted between them