Conclusion

• Developed a State Search Planner by combining two dominant approaches

• Used the planning graph for deriving very effective heuristics to guide backward state search

  • Showed that planning graph heuristics take into account better sub-goals interactions, being more effective

• Provided attractive approaches for trading cost for quality in the heuristic computation

• Result: AltAlt , a very competitive planner

  • Among top three planners according to our experiments.

Previous slide

Next slide

Back to first slide

View graphic version