在部分观测环境下的不确定动作模型学习

近年来,动作模型学习引起了研究人员的极大兴趣.可是,尽管不确定规划已经研究了十几年,动作模型学习的研究仍然集中于经典的确定性动作模型上.提出了在部分观测环境下学习不确定动作模型的算法,该算法可应用于假定人们对转移系统一无所知的情形下进行,输入只有动作-观测序列.在现实世界中,这样的场景很常见.致力于动作是由简单逻辑结构组成的、且观测以一定频率出现的一类问题的研究.学习过程分为3个步骤：首先,计算命题在状态中成立的概率;然后,将命题抽取成效果模式,再抽取前提;最后,对效果模式进行聚类以去除冗余.在基准领域上进行的实验结果表明,动作模型学习技术可推广到不确定的部分观测环境中.     

The rewriting logic semantics project

Rewriting logic is a flexible and expressive logical framework that unifies algebraic denotational semantics and structural operational semantics (SOS) in a novel way, avoiding their respective limitations and allowing succinct semantic definitions. The fact that a rewrite logic theory’s axioms include both equations and rewrite rules provides a useful “abstraction dial” to find the right balance between abstraction and computational observability in semantic definitions. Such semantic definitions are directly executable as interpreters in a rewriting logic language such as Maude, whose generic formal tools can be used to endow those interpreters with powerful program analysis capabilities.     

独立于设计者的行动推理

高水平的智能机器人要求能够独立地对环境进行感知并进行正确的行动推理.在情境演算行动理论中表示带有感知行动及知识的行动推理需要外部设计者为agent写出背景公理、感知结果及相应的知识变化,这是一种依赖于设计者的行动推理.情境演算行动理论被适当扩充,感知器的表示被添加到行动理论的形式语言中,并把agent新知识的产生建立在感知器的应用结果之上.扩充后的系统能够形式化地表示机器人对环境的感知并把感知结果转换为知识,还能进行独立于设计者的行动推理,同时让感知行动的"黑箱"过程清晰化.     

部分观测环境下的时态动作模型学习方法

已有的动作模型学习方法针对确定的或不确定的瞬时动作，而未考虑动作模型中的时态关系。提出了在部分观测环境下自动学习时态动作模型的方法。设计了学习动作持续时间表达式一般形式的两阶段线性回归方法。通过分析命题时间戳设计了动作前提、效果与动作之间时态关系算子的构建算法。在“国际智能规划竞赛”的规划问题集上进行了实验，结果表明了该方法的有效性。     

Logical/linear operators for image curves

We propose a language for designing image measurement operators suitable for early vision. We refer to them as logical/linear (L/L) operators, since they unify aspects of linear operator theory and Boolean logic. A family of these operators appropriate for measuring the low-order differential structure of image curves is developed. These L/L operators are derived by decomposing a linear model into logical components to ensure that certain structural preconditions for the existence of an image curve are upheld. Tangential conditions guarantee continuity, while normal conditions select and categorize contrast profiles. The resulting operators allow for coarse measurement of curvilinear differential structure (orientation and curvature) while successfully segregating edge-and line-like features. By thus reducing the incidence of false-positive responses, these operators are a substantial improvement over (thresholded) linear operators which attempt to resolve the same class of features     

REASONING WITH BACKGROUND KNOWLEDGE–A THREE-LEVEL THEORY

We address the problem of reasoning in cases when knowledge bases containing background knowledge are understood not as sets of formulas (rules and facts) but as collections of partially ordered theories. In our system, the usual, two-part logical structures, consisting of a metalevel and an object level, are augmented by a third level–a referential level. The referential level is a partially ordered collection of theories; it encodes background knowledge. As usual, current situations are described on the object level, and the metalevel is a place for rules that can eliminate some of the models permitted by the object level and the referential level. As a logic of reasoning the system generalizes the standard model of a rational agent: deducing actions and deriving new information about the world from a logical theory–its knowledge base. It is a natural logical system in which priorities on the possible readings of predicates, not special rules of inference, are the main source of nonmonotonicity. We introduce a theory forming operator PT(x) to exploit the priorities, and we investigate its basic logical properties. Then we show how such a system can be augmented by metarules. Although this paper concentrates on basic logical properties of the new theory, this formalism has already been applied to model a number of natural language phenomena such as the notion of text coherence, Gricean maxims, vagueness, and a few others. The paper also briefly compares it with the model of background knowledge of CYC, as proposed by Lenat and Guha.     

Probabilistic Situation Calculus

In this article we propose a Probabilistic Situation Calculus logical language to represent and reason with knowledge about dynamic worlds in which actions have uncertain effects. Uncertain effects are modeled by dividing an action into two subparts: a deterministic (agent produced) input and a probabilistic reaction (produced by nature). We assume that the probabilities of the reactions have known distributions.Our logical language is an extension to Situation Calculae in the style proposed by Raymond Reiter. There are three aspects to this work. First, we extend the language in order to accommodate the necessary distinctions (e.g., the separation of actions into inputs and reactions). Second, we develop the notion of Randomly Reactive Automata in order to specify the semantics of our Probabilistic Situation Calculus. Finally, we develop a reasoning system in MATHEMATICA capable of performing temporal projection in the Probabilistic Situation Calculus.     

A dynamic logic of action

The paper presents a logical treatment of actions based on dynamic logic. This approach makes it possible to reflect clearly the differences between static and dynamic elements of the world, a distinction which seems crucial to us for a representation of actions.Starting from propositional dynamic logic a formal system (DLA) is developed, the programs of which are used to model action types. Some special features of this system are: Basic aspects of time are incorporated in DLA as far as they are needed for our purpose. Names for states and for instants are simulated by formulas. It is possible to express formally that a formula is satisfiable or valid. A special program is introduced to reflect developments which are not caused by an official agent but by external influences.Having presented our basic system DLA we give some examples to illustrate how a logical system of this kind could be used for analysing essential aspects of actions. We therefore touch on such topics asthe results of actions, abilities of the agent, parallel performances of actions. Possibilities of and problems with logical representations of these features are informally discussed. Afterwards first steps towards integrating them into our basic systems are proposed by formally presenting an extension of DLA for each of the topics mentioned. Statement of exclusive submission. This paper has not been submitted elsewhere in identical or similar form.     

Discovering relational and numerical expressions from plan traces for learning action models

In this paper, we propose a domain learning process build on a machine learning-based process that, starting from plan traces with (partially known) intermediate states, returns a planning domain with numeric predicates, and expressive logical/arithmetic relations between domain predicates written in the planning domain definition language (PDDL). The novelty of our approach is that it can discover relations with little information about the ontology of the target domain to be learned. This is achieved by applying a selection of preprocessing, regression, and classification techniques to infer information from the input plan traces. These techniques are used to prepare the planning data, discover relational/numeric expressions, or extract the preconditions and effects of the domain’s actions. Our solution was evaluated using several metrics from the literature, taking as experimental data plan traces obtained from several domains from the International Planning Competition. The experiments demonstrate that our proposal—even with high levels of incompleteness—correctly learns a wide variety of domains discovering relational/arithmetic expressions, showing F-Score values above 0.85 and obtaining valid domains in most of the experiments.

     

Manifold topological multi-resolution analysis method

In this paper, two significant weaknesses of locally linear embedding (LLE) applied to computer vision are addressed: “intrinsic dimension” and “eigenvector meanings”. “Topological embedding” and “multi-resolution nonlinearity capture” are introduced based on mathematical analysis of topological manifolds and LLE. The manifold topological analysis (MTA) method is described and is based on “topological embedding”. MTA is a more robust method to determine the “intrinsic dimension” of a manifold with typical topology, which is important for tracking and perception understanding. The manifold multi-resolution analysis (MMA) method is based on “multi-resolution nonlinearity capture”. MMA defines LLE eigenvectors as features for pattern recognition and dimension reduction. Both MTA and MMA are proved mathematically, and several examples are provided. Applications in 3D object recognition and 3D object viewpoint space partitioning are also described.