Simple and Fast Strong Cyclic Planning for Fully-Observable Nondeterministic Planning Problems

Evaluation Goal: Evaluate FIP, which implements the Basic algorithm together with our two extensions, on problem instances from 4 domains in the IPC2008 FOND track Blocksworld, faults, first-responders, forest Compared against two state-of-the-art planners: Gamer & MBP Results and Discussion FIP has a better problem coverage than Gamer & MBP [Table 1] Gamer & MBP cannot solve more than 10 problems in Blocksworld FIP can solve all problems efficiently (cutoff time 1,200 seconds) FIP outperforms other planners w.r.t. CPU time t (expressed in seconds) and solution size s (expressed in the number of states in the solution policy) [Table 2] Simple and Fast Strong Cyclic Planning for FullyObservable Nondeterministic Planning Problems