We describe and evaluate the algorithmic techniques that are used in the FF planning system. Like the HSP system, FF relies on forward state space search, using a heuristic that estimates goal distances by ignoring delete lists. Unlike HSP's heuristic, our method does not assume facts to be independent. We introduce a novel search strategy that combines Hill-climbing with systematic search, and we show how other powerful heuristic information can be extracted and used to prune the search space. FF was the most successful automatic planner at the recent AIPS-2000 planning competition. We review the results of the competition, give data for other benchmark domains, and investigate the reasons for the runtime performance of FF compared to HSP.

Over the years increasingly sophisticated planning algorithms have been developed. These have made for more efficient planners, but unfortunately these planners still suffer from combinatorial complexity even in simple domains. Theoretical results demonstrate that planning is in the worst case intractable. Nevertheless, planning in particular domains can often be made tractable by utilizing additional domain structure. In fact, it has long been acknowledged that domain independent planners need domain dependent information to help them plan effectively. In this

The Blackbox planning system unifies the planning as satisfiability framework (Kautz and Selman 1992, 1996) with the plan graph approach to STRIPS planning (Blum and Furst 1995). We show that STRIPS problems can be directly translated into SAT and efficiently solved using new randomized systematic solvers. For certain computationally challenging benchmark problems this unified approach outperforms both SATPLAN and Graphplan alone. We also demonstrate that polynomialtime SAT simplification algorithms applied to the encoded problem instances are a powerful complement to the "mutex" propagation algorithm that works directly on the plan graph. 1 Introduction It has often been observed that the classical AI planning problem (that is, planning with complete and certain information) is a form of logical deduction. Because early attempts to use general theorem provers to solve planning problems proved impractical, research became focused on specialized planning algorithms. Sometimes the rela...

In the recent AIPS98 Planning Competition, the hsp planner, based on a forward state search and a domain-independent heuristic, showed that heuristic search planners can be competitive with state of the art Graphplan and Satisfiability planners. hsp solved more problems than the other planners but it often took more time or produced longer plans. The main bottleneck in hsp is the computation of the heuristic for every new state. This computation may take up to 85% of the processing time. In this paper, we present a solution to this problem that uses a simple change in the direction of the search. The new planner, that we call hspr, is based on the same ideas and heuristic as hsp, but searches backward from the goal rather than forward from the initial state. This allows hspr to compute the heuristic estimates only once. As a result, hspr can produce better plans, often in less time. For example, hspr solves each of the 30 logistics problems from Kautz and Selman in less than 3 seconds. This is two orders of magnitude faster than blackbox. At the same time

If an agent does not have complete information about the world-state, it must reason about alternative possible states of the world and consider whether any of its actions can reduce the uncertainty. Agents controlled by a contingent planner seek to generate a robust plan, that accounts for and handles all eventualities, in advance of execution. Thus a contingent plan may include sensing actions which gather information that is later used to select between different plan branches. Unfortunately, previous contingent planners suffered defects such as confused semantics, incompleteness, and inefficiency. In this paper we describe SGP, a descendant of Graphplan that solves contingent planning problems. SGP distinguishes between actions that sense the value of an unknown proposition from those that change its value. SGP does not suffer from the forms of incompleteness displayed by CNLP and Cassandra. Furthermore, SGP is relatively fast. 1 Introduction Classical planners make the unrealisti...

Many planning domains require a richer notion of time in which actions can overlap and have different durations. The key to fast performance in classical planners (e.g., Graphplan, ipp, and Blackbox) has been the use of a disjunctive representation with powerful mutual exclusion reasoning. This paper presents TGP a new algorithm for temporal planning. TGP operates

The past five years have seen dramatic advances in planning algorithms, with an emphasis on propositional methods such as Graphplan and compilers that convert planning problems into propositional CNF formulae for solution via systematic or stochastic SAT methods. Related work on the Deep Space One spacecraft control algorithms advances our understanding of interleaved planning and execution. In this survey,we explain the latest techniques and suggest areas for future research.

. This paper outlines the basic principles underlying reasoning about resources in IPP, which is a classical planner based on planning graphs originally introduced with the graphplan system. The main idea is to deal with resources in a strictly action-centered way, i.e., one specifies how each action consumes or produces resources, but no explicit temporal model is used. This avoids the computational problems of solving general constraint satisfaction problems by using instead interval arithmetics and propagation of resource requirements over time steps in the planning graph. 1 Actions that provide, produce, and consume Resources The starting point for the language extension is the ADL subset that is available in IPP 3.0 [7]. It offers universally quantified and conditional effects, atomic negation, equality as well as quantified and conditional goals. To reason about resources, an action description is extended in the following way: 1. Following the "ordinary preconditions" (which a...

There has been a great deal of recent work on new approaches to efficiently generating plans in systems such as Graphplan and SATplan. However, these systems only provide an impoverished representation language compared to other planners, such as UCPOP, ADL, or Prodigy. This makes it difficult to represent planning problems using these new planners. This paper addresses this problem by providing a completely automated set of transformations for converting a UCPOP domain representation into a Graphplan representation. The set of transformations extends the Graphplan representation language to include disjunctions, negations, universal quantification, conditional effects, and axioms. We tested the resulting planner on the 18 test domains and 41 problems that come with the UCPOP 4.0 distribution. Graphplan with the new preprocessor is able to solve every problem in the test set and on the hard problems (i.e., those that require more than one second of CPU time) it can solve them significantly faster than UCPOP. While UCPOP was unable to solve 7 of the test problems within a search limit of 100,000 nodes (which requires 414 to 980 CPU seconds), Graphplan with the preprocessor solved them all in under 15 CPU seconds (including the preprocessing time).

We describe some new preprocessing techniques that enable faster domain-independent planning. The first set of techniques is aimed at inferring state constraints from the structure of planning operators and the initial state. Our methods consist of generating hypothetical state constraints by inspection of operator effects and preconditions, and checking each hypothesis against all operators and the initial conditions. Another technique extracts (supersets of) predicate domains from sets of ground literals obtained by Graphplan-like forward propagation from the initial state. Our various techniques are implemented in a package called DISCOPLAN. We show preliminary results on the effectiveness of adding computed state constraints and predicate domains to the specification of problems for SAT-based planners such as SATPLAN or MEDIC. The results suggest that large speedups in planning can be obtained by such automated methods, potentially obviating the need for adding h...