Product Representation of Belief Spaces in Planning under Partial Observability
نویسنده
چکیده
We present a product representation of belief spaces for planning under partial observability. In earlier work we investigated backward plan construction based on a combination operation for belief states. The main problem in explicit construction of belief states is their high number. To remedy this problem, we refrain from representing individual belief states explicitly, and instead represent parts of the belief space in a factored form. The factorization is induced by the division of the state space to observational classes, each consisting of observationally indistinguishable states. Finally, we show that the representation leads to a simple planning algorithm that is competitive with other algorithms for planning under partial observability.
منابع مشابه
LTLf and LDLf Synthesis under Partial Observability
In this paper, we study synthesis under partial observability for logical specifications over finite traces expressed in LTLf /LDLf . This form of synthesis can be seen as a generalization of planning under partial observability in nondeterministic domains, which is known to be 2EXPTIMEcomplete. We start by showing that the usual “belief-state construction” used in planning under partial observ...
متن کامل($phi,rho$)-Representation of $Gamma$-So-Rings
A $Gamma$-so-ring is a structure possessing a natural partial ordering, an infinitary partial addition and a ternary multiplication, subject to a set of axioms. The partial functions under disjoint-domain sums and functional composition is a $Gamma$-so-ring. In this paper we introduce the notions of subdirect product and $(phi,rho)$-product of $Gamma$-so-rings and study $(phi,rho)$-represen...
متن کاملPattern-Database Heuristics for Partially Observable Nondeterministic Planning
Heuristic search is the dominant approach to classical planning. However, many realistic problems violate classical assumptions such as determinism of action outcomes or full observability. In this paper, we investigate how – and how successfully – a particular classical technique, namely informed search using an abstraction heuristic, can be transferred to nondeterministic planning under parti...
متن کاملPOND: The Partially-Observable and Non-Deterministic Planner
This paper describes POND, a planner developed to solve problems characterized by partial observability and nondeterminism. POND searches in the space of belief states, guided by a relaxed plan heuristic. Many of the more interesting theoretical issues showcased by POND show up within its relaxed plan heuristics. Namely, the exciting topics are defining distance estimates between belief states,...
متن کاملSummary of Experiments in Belief-Space Planning at the Laboratory for Perceptual Robotics
In robotics, observations and actions are rarely certain, making it difficult to plan and take actions to reach goal states. Belief-space planners are often used to overcome the uncertainty and partial observability, but make it difficult to define generic tasks. In recent work, we presented a planner and task representation framework capable of alleviating these problems. This paper is a summa...
متن کامل