The current best conformant probabilistic planners encode the problem as a bounded length CSP or SAT problem. While these approaches can find optimal solutions for given plan lengths, they often do not scale for large problems or plan lengths. As has been shown in classical planning, heuristic search outperforms CSP/SAT techniques (especially when a plan length is not given a priori). The problem with applying heuristic search in probabilistic planning is that effective heuristics are as yet lacking. In this work, we apply heuristic search to conformant probabilistic planning by adapting planning graph heuristics developed for non-deterministic planning. We evaluate a straight-forward application of these planning graph techniques, which amounts to exactly computing the distribution over reachable relaxed planning graph layers. Computing these distributions is costly, so we apply Sequential Monte Carlo to approximate them. We demonstrate on several domains how our approach enables our planner to far out-scale existing (optimal) probabilistic planners and still find reasonable quality solutions.

Focusing on the computation of conformant plans whose verification can be done efficiently, we have recently proposed a polynomial scheme for mapping conformant problems P with deterministic actions into classical problems K(P). The scheme is sound as the classical plans are all conformant, but is incomplete as the converse relation does not always hold. In this paper, we extend this work and consider an alternative, more powerful translation based on the introduction of epistemic tagged literals KL/t where L is a literal in P and t is a set of literals in P unknown in the initial situation. The translation ensures that a plan makes KL/t true only when the plan makes L certain in P given the assumption that t is initially true. We show that under general conditions the new translation scheme is complete and that its complexity can be characterized in terms of a parameter of the problem that we call conformant width. We show that the complexity of the translation is exponential in the problem width only, find that the width of almost all benchmarks is 1, and show that a conformant planner based on this translation solves some interesting domains that cannot be solved by other planners. This translation is the basis for T0, the best performing planner

Conformant planning is the problem of finding a sequence of actions for achieving a goal in the presence of uncertainty in the initial state or action effects. The problem has been approached as a path-finding problem in belief space where good belief representations and heuristics are critical for scaling up. In this work, a different formulation is introduced for conformant problems with deterministic actions where they are automatically converted into classical ones and solved by an off-the-shelf classical planner. The translation maps literals L and sets of assumptions t about the initial situation, into new literals KL/t that represent that L must be true if t is initially true. We lay out a general translation scheme that is sound and establish the conditions under which the translation is also complete. We show that the complexity of the complete translation is exponential in a parameter of the problem called the conformant width, which for most benchmarks is bounded. The planner based on this translation exhibits good performance in comparison with existing planners, and is the basis for T0, the best performing planner in the Conformant Track of the 2006 International Planning Competition. 1.

The problem of planning in the presence of sensing has been addressed in recent years as a nondeterministic search problem in belief space. In this work, we use ideas advanced recently for compiling conformant problems into classical ones for introducing a different approach where contingent problems P are mapped into non-deterministic problems X(P) in state space. We also identify a contingent width parameter, and show that for problems P with bounded contingent width, the translation is sound, polynomial, and complete. We then solve X(P) by using a relaxation X + (P) that is a classical planning problem. The formulation is tested experimentally over contingent benchmarks where it is shown to yield a planner that scales up better than existing contingent planners. 1

Abstract not found

Planning in belief space with a Labelled Uncertainty Graph, LUG, is an approach that uses a very compact planning graph to guide search in the space of belief states to construct conformant and contingent plans. A conformant plan is a plan that transitions (without sensing) all possible initial states through possibly non-deterministic actions to a goal state. A contingent plan adds the ability to observe state variables and branch execution. The LUG

A number of today’s state-of-the-art planners are based on forward state-space search. The impressive performance can be attributed to progress in computing domain independent heuristics that perform well across many domains. However, it is easy to find domains where such heuristics provide poor guidance, leading to planning failure. Motivated by such failures, the focus of this paper is to investigate mechanisms for learning domain-specific knowledge to better control forward search in a given domain. While there has been a large body of work on inductive learning of control knowledge for AI planning, there is a void of work aimed at forward-state-space search. One reason for this may be that it is challenging to specify a knowledge representation for compactly representing important concepts across a wide range of domains. One of the main contributions of this work is to introduce a novel feature space for representing such control knowledge. The key idea is to define features in terms of information computed via relaxed plan extraction, which has been a major source of success for non-learning planners. This gives a new way of leveraging relaxed planning techniques in the context of learning. Using this feature space, we describe three forms of control knowledge—reactive policies (decision list rules and measures of progress) and linear heuristics—and show how to learn them and incorporate them into forward state-space search. Our empirical results show that our approaches are able to surpass state-of-the-art nonlearning planners across a wide range of planning competition domains.

We present a new approach for finding generalized contingent plans with loops and branches in situations where there is uncertainty in state properties and object quantities, but lack of probabilistic information about these uncertainties. We use a state abstraction technique from static analysis of programs, which uses 3-valued logic to compactly represent belief states with unbounded numbers of objects. Our approach for finding plans is to incrementally generalize and merge input example plans which can be generated by classical planners. The expressiveness and scope of this approach are demonstrated using experimental results on common benchmark domains.

Thanks to recent advances, AI Planning has become the underlying technique for several applications. Amongst these, a prominent one is automated Web Service Composition (WSC). One important issue in this context has been hardly addressed so far: WSC requires dealing with background ontologies. The support for those is severely limited in current planning tools. We introduce a planning formalism that faithfully represents WSC. We show that, unsurprisingly, planning in such a formalism is very hard. We then identify an interesting special case that covers many relevant WSC scenarios, and where the semantics are simpler and easier to deal with. This opens the way to the development of effective support tools for WSC. Furthermore, we show that if one additionally limits the amount and form of outputs that can be generated, then the set of possible states becomes static, and can be modelled in terms of a standard notion of initial state uncertainty. For this, effective tools exist; these can realize scalable WSC with powerful background ontologies. In an initial experiment, we show how scaling WSC instances are comfortably solved by a tool incorporating modern planning heuristics.

The primary revolution in automated planning in the last decade has been the very impressive scaleup in planner performance. A large part of the credit for this can be attributed squarely to the invention and deployment of powerful reachability heuristics. Most, if not all, modern reachability heuristics are based on a remarkably extensible data structure called the planning graph, which made its debut as a bit player in the success of GraphPlan, but quickly grew in prominence to occupy the center stage. Planning graphs are a cheap means to obtain informative look-ahead heuristics for search and have become ubiquitous in state of the art heuristic search planners. We present the foundations of planning graph heuristics in classical planning and explain how their flexibility lets them adapt to more expressive scenarios that consider action costs, goal utility, numeric resources, time, and uncertainty.