Exploiting state constraints in heuristic state-space planning
Ioannis Vlahavas2000
Sign up for access to the world's latest research
checkGet notified about relevant papers
checkSave papers to use in your research
checkJoin the discussion with peers
checkTrack your impact
Abstract
In the last years, some very promising domain independent heuristic state-space planners for STRIPS worlds, like ASP/HSP, HSPr and GRT, have been presented. These planners achieve remarkable performance in some domains, like the blocks world, the logistics and the gripper, but they are not effective in other domains, like the grid and the mystery. In this paper we propose the use of state constraints in heuristic state space planning. We claim that one of the causes for the pre-mentioned failures is the absence of domain specific knowledge about properties that characterize every valid and complete state. We propose the inclusion of state constraints in the domain definition and we present how they can be exploited by heuristic planners in order to decompose a problem into subproblems that are easily solvable. We give performance results that exhibit significant speedup in the problem solving process. Finally, we give a notion of how problem decomposition can accelerate other planners, like GRAPHPLAN and BLACKBOX.
Related papers
The GRT planning system: Backward heuristic construction in forward state-space planning
2011
This paper presents GRT, a domain-independent heuristic planning system for STRIPS worlds. GRT solves problems in two phases. In the pre-processing phase, it estimates the distance between each fact and the goals of the problem, in a backward direction. Then, in the search phase, these estimates are used in order to further estimate the distance between each intermediate state and the goals, guiding so the search process in a forward direction and on a best-first basis. The paper presents the benefits from the adoption of opposite directions between the preprocessing and the search phases, discusses some difficulties that arise in the pre-processing phase and introduces techniques to cope with them. Moreover, it presents several methods of improving the efficiency of the heuristic, by enriching the representation and by reducing the size of the problem. Finally, a method of overcoming local optimal states, based on domain axioms, is proposed. According to it, difficult problems are decomposed into easier sub-problems that have to be solved sequentially. The performance results from various domains, including those of the recent planning competitions, show that GRT is among the fastest planners.
AltAlt: Combining the Advantages of Graphplan and Heuristic State Search
2000
Most recent strides in scaling up planning have centered around two competing themes-disjunctive planners, exemplified by Graphplan, and heuristic state search planners, exemplified by HSP and HSP-R. In this paper, we describe a planner called AltAlt, which successfully combines the advantages of the two competing paradigms to develop a planner that is significantly more powerful than either of the approaches. AltAlt uses Graphplan's planning graph in a novel manner to derive very effective search heuristics which are then used to drive a heuristic state search planner. Al-tAlt is implemented by splicing together implementations of STAN, a state-of-the-art Graphplan implementation, and HSP-r, a heuristic search planner. We present empirical results in a variety of domains that show the significant scale-up power of our combined approach. We will also present a variety of possible optimizations for our approach, and discuss the rich connections between our work and the literature on state-space search heuristics.
Discovering state constraints in DISCOPLAN: Some new results
PROCEEDINGS OF THE NATIONAL …, 2000
DISCOPLAN is an implemented set of efficient preplanning algorithms intended to enable faster domain-independent planning. It includes algorithms for discovering state constraints (invariants) that have been shown to be very useful, for example, for speeding up SAT-based planning. DISCOPLAN originally discovered only certain types of implicative constraints involving up to two fluent literals and any number of static literals, where one of the fluent literals contains all of the variables occurring in the other literals; only planning domains with STRIPS-like operators were handled. We have now extended DISCOPLAN in several directions. We describe new techniques that handle operators with conditional effects, and enable discovery of several new types of constraints. Moreover, discovered constraints can be fed back into the discovery process to obtain additional constraints. Finally, we outline unimplemented (but provably correct) methods for discovering additional types of constraints, including constraints involving arbitrarily many fluent literals. 0 -valuedness) constraints, and constraints involving ar-From: AAAI-00 Proceedings.
Planning graph as the basis for deriving heuristics for plan synthesis by state space and CSP search
Artificial Intelligence, 2002
Most recent strides in scaling up planning have centered around two competing themesdisjunctive planners, exemplified by Graphplan, and heuristic state search planners, exemplified by UNPOP, HSP and HSP-r. In this paper, we present a novel approach for successfully harnessing the advantages of the two competing paradigms to develop planners that are significantly more powerful than either of the approaches. Specifically, we show that the polynomial-time planning graph structure that the Graphplan builds provides a rich substrate for deriving a family of highly effective heuristics for guiding state space search as well as CSP style search. The main leverage provided by the planning graph structure is a systematic and graded way to take subgoal interactions into account in designing state space heuristics. For state space search, we develop several families of heuristics, some aimed at search speed and others at optimality of solutions, and analyze many approaches for improving the cost-quality tradeoffs offered by these heuristics. Our normalized empirical comparisons show that our heuristics handily outperform the existing state space heuristics. For CSP style search, we describe a novel way of using the planning graph structure to derive highly effective variable and value ordering heuristics. We show that these heuristics can be used to improve Graphplan's own backward search significantly. To demonstrate the effectiveness of our approach vis a vis the state-of-the-art in plan synthesis, we present AltAlt, a planner literally cobbled together from the implementations of Graphplan and state search style planners using our theory. We evaluate AltAlt on the suite of problems used in the recent AIPS-2000 planning competition. The results place AltAlt in the top tier of the competition planners-outperforming both Graphplan based and heuristic search based planners. 2001 Published by Elsevier Science B.V. ✩ Preliminary versions of parts of this work have been presented at AAAI: S 0 0 0 4 -3 7 0 2 ( 0 1 ) 0 0 1 5 8 -8 74 X. Nguyen et al. / Artificial Intelligence 135 (2002)
Heuristic planning for PDDL+ domains
International Joint Conference on Artificial Intelligence, 2016
Planning with hybrid domains modelled in PDDL+ has been gaining research interest in the Automated Planning community in recent years. Hybrid domain models capture a more accurate representation of real world problems that involve continuous processes than is possible using discrete systems. However, solving problems represented as PDDL+ domains is very challenging due to the construction of complex system dynamics, including non-linear processes and events. In this paper we introduce DiNo, a new planner capable of tackling complex problems with non-linear system dynamics governing the continuous evolution of states. DiNo is based on the discretise-and-validate approach and uses the novel Staged Relaxed Planning Graph+ (SRPG+) heuristic, which is introduced in this paper. Although several planners have been developed to work with subsets of PDDL+ features, or restricted forms of processes, DiNo is currently the only heuristic planner capable of handling non-linear system dynamics combined with the full PDDL+ feature set.
An evolutionary metaheuristic based on state decomposition for domain-independent satisficing planning
2010
DAEX is a metaheuristic designed to improve the plan quality and the scalability of an encapsulated planning system. DAEX is based on a state decomposition strategy, driven by an evolutionary algorithm, which benefits from the use of a classical planning heuristic to maintain an ordering of atoms within the individuals. The proof of concept is achieved by embedding the domain-independent satisficing YAHSP planner and using the critical path h 1 heuristic. Experiments with the resulting algorithm are performed on a selection of IPC benchmarks from classical, cost-based and temporal domains. Under the experimental conditions of the IPC, and in particular with a universal parameter setting common to all domains, DAEYAHSP is compared to the best planner for each type of domain. Results show that DAEYAHSP performs very well both on coverage and quality metrics. It is particularly noticeable that DAEX improves a lot on plan quality when compared to YAHSP, which is known to provide largely suboptimal solutions, making it competitive with state-of-the-art planners. This article gives a full account of the algorithm, reports on the experiments and provides some insights on the algorithm behavior.
Towards learning domain-independent planning heuristics
arXiv (Cornell University), 2017
Automated planning remains one of the most general paradigms in Artificial Intelligence, providing means of solving problems coming from a wide variety of domains. One of the key factors restricting the applicability of planning is its computational complexity resulting from exponentially large search spaces. Heuristic approaches are necessary to solve all but the simplest problems. In this work, we explore the possibility of obtaining domain-independent heuristic functions using machine learning. This is a part of a wider research program whose objective is to improve practical applicability of planning in systems for which the planning domains evolve at run time. The challenge is therefore the learning of (corrections of) domain-independent heuristics that can be reused across different planning domains.
Exploiting Domain Knowledge with a Concurrent Hierarchical Planner
Based on recent research about coordinating concurrent hierarchical plans (CHiPs), weintroduce a sound and complete hierarchical planner that can better reason about precomputed conditions (summary information) of abstract plans to potentially make better refinement decisions than previous approaches. A reasonable criticism of this technique is that the summary information can grow exponentially as it is propagated up a plan hierarchy. This paper analyzes the complexityofthe problem to show that in spite of this exponential growth, finding solutions at higher levels is still exponentially cheaper than at lower levels. In addition, this paper offers heuristics, including "fewest threats first" (FTF) and "expand most threats first" (EMTF), that take advantage of summary information to smartly direct the search for a global plan. Experiments show that for a particular domain these heuristics could greatly improve the search for the optimal global plan compared to tw...
DISCOPLAN: an efficient on-line system for computing planning domain invariants
DISCOPLAN is an efficient system for discovering state invariants in planning domains with conditional effects. Among the types of invariants found are implicative constraints relating a fluent predication to a fluent or static predication (with allowance for static supplementary conditions), single-valuedness constraints, exclusiveness constraints, and several others. The algorithms used are polynomial-time for any fixed bound on the number of literals in an invariant. Some combinations of constraints are found by simultaneous induction, and the methods can be iterated by expanding operators using previously found invariants. The invariants found by DISCOPLAN have been shown to enable large performance gains in SAT planners, and they can also be helpful in planning domain development and debugging.
Related topics
Related papers
HPP: A Heuristic Progressive Planner
In the last decade three main ideas have revolutionized the Artificial Intelligence (AI) planning methods: the plan graph search structure, planning as satisfiability and the search guided by a heuristic function calculated from a relaxed plan. In this paper we present a new planner, named HPP (Heuristic Progressive Planner), that has been implemented using several ideas extracted from other heuristic planners. However, HPP includes a new module (analysis reachability module) that is able to exclude irrelevant domain-dependent operators for the planning process. The analysis performed by this module avoids to expand several parts of the search tree allowing to solve more problems. Finally, the HPP planner has been incorporated into a distributed multi-agent system to allow exploring in parallel subsets of the search space.
Multiobjective heuristic state-space planning
Artificial Intelligence, 2003
Modern domain-independent heuristic planners evaluate their plans on the single basis of their length. However, in real-world problems, there are other criteria that also play an important role, e.g., resource consumption, profit, safety, etc. This paper enhances the GRT planner, an efficient domain-independent heuristic state-space planner, with the ability to consider multiple criteria. The GRT heuristic is based on the estimation of the distances between each fact of a problem and the goals. The new planner, called MO-GRT, uses a weighted A * strategy and a multiobjective heuristic function, computed over a weighted hierarchy of user-defined criteria. Its computation is based on sets of non-dominated cost-vectors assigned to the problem facts, which estimate the total cost of achieving the facts from the goals, using alternative paths. Experiments show that a change in the criteria weights or scales affects both the quality of the resulting plan and the planning time. The proposed approach can easily be adapted to other modern heuristic state-space planners.
Domain-Dependent Heuristics and Tie-Breakers: Topics in Automated Planning
Revista Eletrônica de Iniciação Científica em Computação, 2018
Automated planning is an important general problem solving technique in Artificial Intelligence. Given an initial state, a goal and a set of operators, we want to find a sequence of operators leading us to the goal. What makes planning interesting is that it can model different domains into planning tasks and solve them using a single method. In this work, we approach two different topics in planning. First, we study heuristics for the airport ground traffic problem and propose new heuristics that are better than any other known method. In the second part, we study tie-breakers for the A* search algorithm. We propose a new tie-breaking method that is proved to be the best possible and also show that our methods solve more instances than previous methods in literature
Universal Classical Planner: An algorithm for unifying State-space and Plan-space planning
1995
We present a plan representation and a generalized algorithm template, called UCP, for unifying the classical plan-space and state-space planning approaches within a single framework. UCP models planning as a process of refining a partial plan. The plan-space and state-space planning approaches are cast as complementary refinement strategies operating on the same partial plan representation. UCP has the freedom to arbitrarily and opportunistically interleave plan-space and state-space refinements within a single planning episode. This allows it reap the benefits of both state-space and plan-space planning approaches. We discuss the coverage, completeness and systematicity of UCP. We also present some preliminary empirical results that demonstrate the utility of combining state-space and plan-space approaches.
State Space Planning Using Transaction Logic
State space planning algorithms have been considered as one of the main classical planning techniques to solve classical planning problems since 1960. In this paper, we show that Transaction Logic is an appropriate language and framework to study and compare these planning algorithms, which enables one to have more efficient planners in logic programming frameworks. Specifically, we take STRIPS planning and forward state space planning algorithms, and show that the specification of these algorithms in Transaction Logic lets one implement complicated planning algorithms in declarative programming languages (e.g. Prolog). We first provide a formal representation of these planning algorithms in Transaction Logic, which can be used to automatically translate STRIPS planning problems in PDDL to Transaction Logic rules. Then, we use the resulting Transaction Logic rules to solve planning problems and compare the performance of those algorithms in our simple interpreter implemented in XSB Prolog. We use several case studies to show how the linear STRIPS planning algorithm is faster than forward state space search. Our experiments highlight the fact that a planner implemented by logic programming framework can become faster if an appropriate planning algorithm is applied.
HYBRID ACE: COMBINING SEARCH DIRECTIONS FOR HEURISTIC PLANNING
Computational Intelligence, 2005
One of the most promising trends in Domain-Independent AI Planning, nowadays, is state-space heuristic planning. The planners of this category construct general but efficient heuristic functions, which are used as a guide to traverse the state space either in a forward or in a backward direction. Although specific problems may favor one or the other direction, there is no clear evidence why any of them should be generally preferred. This paper presents Hybrid-AcE, a domain-independent planning system that combines search in both directions utilizing a complex criterion that monitors the progress of the search, to switch between them. Hybrid AcE embodies two powerful domain-independent heuristic functions extending one of the AcE planning systems. Moreover, the system is equipped with a fact-ordering technique and two methods for problem simplification that limit the search space and guide the algorithm to the most promising states. The bi-directional system has been tested on a variety of problems adopted from the AIPS planning competitions with quite promising results.
Domain-independent construction of pattern database heuristics for cost-optimal planning
2007
Heuristic search is a leading approach to domain-independent planning. For cost-optimal planning, however, existing admissible heuristics are generally too weak to effectively guide the search. Pattern database heuristics (PDBs), which are based on abstractions of the search space, are currently one of the most promising approaches to developing better admissible heuristics. The informedness of PDB heuristics depends crucially on the selection of appropriate abstractions (patterns). Although PDBs have been applied to many search problems, including planning, there are not many insights into how to select good patterns, even manually. What constitutes a good pattern depends on the problem domain, making the task even more difficult for domain-independent planning, where the process needs to be completely automatic and general. We present a novel way of constructing good patterns automatically from the specification of planning problem instances. We demonstrate that this allows a domainindependent planner to solve planning problems optimally in some very challenging domains, including a STRIPS formulation of the Sokoban puzzle.
Encoding Planning Constraints into Partial Order Planning Domains
1998
This paper presents a technique for encoding the representation of a constrained planning problem through the generation of an equivalent unconstrained planning problem. In many real situations the initial state/goals characterization of a planning problem is not satisfactory, this motivates the notion of a constrained planning problem, i.e. a planning problem for which the user speci es additional constraints on the problem solution in order to: give a limit to the length of the solution plan, use or avoid speci c action instances, use particular strategies in adding steps or precedence constraints to the plan, reach some intermediate states. A common way to implement constrained planning is modifying the existing planner in order to take into account of the user additional constraints. In this paper we present a planner independent approach, which operates on the problem and constraints representation instead of modifying the planning algorithm. A planning constraint description language (PCDL) is therefore introduced. It is shown that a signi cant subset of PCDL can be encoded by modifying (domain preprocessing phase) the planning domain and goals. The original problem can then be solved by submitting the modi ed problem to the planner. The solutions to the modi ed problem de ned by the new domain correspond (solution postprocessing phase) to the solutions of the original constrained problem. A signi cant result is that domain preprocessing phase has a linear time/space cost in the domain and additional constraints dimensions, and the solution postprocessing phase has also linear cost on the solution length. A the-oretical consequence of the equivalence result is that representation of constraints and strategies does not require a planning model which is more powerful than an ordinary partial order planning.
Plan aggregation for strong cyclic planning in nondeterministic domains
Artificial Intelligence, 2014
We describe a planning algorithm, NDP2, that finds strong-cyclic solutions to nondeterministic planning problems by using a classical planner to solve a sequence of classical planning problems. NDP2 is provably correct, and fixes several problems with prior work.
Hybrid planning for partially hierarchical domains
1998
Hierarchical task network and action-based planning approaches have traditionally been studied separately. In many domains, human expertise in the form of hierarchical reduction schemas exists, but is incomplete. In such domains, hybrid approaches that use both HTN and action-based planning techniques are needed. In this paper, we extend our previous work on refinement planning to include hierarchical planning. Specifically, we provide a generalized plan-space refinement that is capable of handling non-primitive actions. The generalization provides a principled way of handling partially hierarchical domains, while preserving systematicity, and respecting the user-intent inherent in the reduction schemas. Our general account also puts into perspective the many surface differences between the HTN and action-based planners, and could support the transfer of progress between HTN and action-based planning approaches.
Loading Preview
Sorry, preview is currently unavailable. You can download the paper by clicking the button above.