Strips Planning Systems -The Sussman Anamoly
- There are some, drawbacks with using recursive STRIPS.
- The drawbacks can be explained with an example. Let us consider an discs-world problem that has to delead using recursive STRIPS.
Fig. The Sussman Anamoly
- The goal state as shown in Fig. 3.10.3 is,
- On(P, Q) ∧ On(Q, R) ∧ On(R, F) ∧ Clear(P) ∧ Clear(F)
- The initial state is,
- On(R, P) ∧ On(Q, F) ∧ Clear(Q) ∧ Clear(R)
- Inorder to achieve goal state using the STRIPS rules one of the conjunct has to be selected.
- Let it selects On(P, Q) first. So it moves R on to floor and then moves P on to Q.
- But the other conjunct is On(Q, R) in order to do this it has to revert the process it has done.
- So, let us assume On(Q, R) is selected first. It performs the operation of moving Q onto R. But the other conjunct On(P, Q) is selected and again this process has to be re-achieved.
- This problem of not having a minimum number of operations to reach the goal state is known as Sussman Anomaly.
- This is the problem due to the usage of depth-first search in recursive STRIPS.
- So we may think of using bread-first search, but, its usage is not feasible in realistic problems.
- A solution to this problem is to use breadth-first, backward-directed search. Using backward search is efficient, because the number of conjuncts in goals state wff are less than those of initial state description.