Empirical Evaluation: Setup

This section presents our implementation of the ${\cal C}{\sl AltAlt}$ and $POND$ planners and the domains we use in the experiments. All tests were run in Linux on an x86 machine with a 2.66GHz P4 processor and 1GB RAM with a timeout of 20 minutes. Both CAltAlt and $POND$ used a heuristic weight of five in the, respective, A* and AO* searches. We compare with the competing approaches (CGP, SGP, GPT v1.40, MBP v0.91, KACMBP, YKA, and CFF) on several domains and problems. Our planners and all domain and problem files for all of the compared planners can be found in the online appendix.