Graph Search and STRIPS Planning
Abstract:
AI is founded upon that belief that intelligence is based on computation. It seems likely that a deep understanding of computation will prove useful in the investigation of intelligence. In any computational process efficiency is desirable. This chapter continues the investigation of techniques for improving the efficiency of computations that search. The theory of computation provides a framework for classifying computations that search, i.e., those computations which can not be done in polynomial time. As noted in earlier chapters, there appears to be a connection between the intuitive classification of search problems and formal complexity classes. Constraint satisfaction corresponds to the complexity class NP. On the other hand, graph search corresponds to the complexity class PSPACE. Graph search problems seem to be quite different from constraint satisfaction problems. This chapter describes the most pragmatically effective known algorithms for solving graph search problems. This chapter also introduces the closely related problem of STRIPS planning along with simple planning algorithms. Just as SAT can be viewed as a prototypical NP complete problem, STRIPS planning is a prototypical PSPACE complete problem. The most pragmatically effective known algorithms for STRIPS planning are called "nonlinear planners " and are discussed in the next chapter.
Citations
| 507 | A note on two problems in connection with graphs – Dijkstra - 1959 |
| 459 | A Formal Basis for the Heuristic Determination of Minimum Cost Paths – Hart, Nilsson, et al. - 1968 |
| 280 | Depth-first iterative-deepening: An optimal admissible tree search – Korf - 1985 |
| 17 | Experiments with the graph traverser program – Doran, Michie - 1966 |
