The notion of time is ubiquitous in any activity that requires intelligence. In particular, several important notions like change, causality, action are described in terms of time. Therefore, the representation of time and reasoning about time is of crucial importance for many Artificial Intelligence systems. Specifically during the last 10 years, it has been attracting the attention of many AI researchers. In this survey, the results of this work are analysed. Firstly, Temporal Reasoning is defined. Then, the most important representational issues which determine a Temporal Reasoning approach are introduced: the logical form on which the approach is based, the ontology (the units taken as primitives, the temporal relations, the algorithms that have been developed,. . . ) and the concepts related with reasoning about action (the representation of change, causality, action,. . . ). For each issue the different choices in the literature are discussed. 1 Introduction The notion of time i...

We analyse the kinematics of probabilistic term weights at retrieval time for di erent Information Retrieval models. We present four models based on di erent notions of probabilistic retrieval. Two of these models are based on classical probability theory and can be considered as prototypes of models long in use in Information Retrieval, like the Vector Space Model and the Probabilistic Model. The two other models are based on a logical technique of evaluating the probability of a conditional called imaging, one is a generalisation of the other. We analyse the transfer of probabilities occurring in the term space at retrieval time for these four models, compare their retrieval performance using classical test collections, and discuss the results. We believe that our results provide useful suggestions on how to improve existing probabilistic models of Information Retrieval by taking into consideration term-term similarity.

We develop an approach to planning with incomplete information that is based on three elements: 1. a high-level language for describing the effects of actions on both the world and the agent's beliefs that we call pomdp theories 2. a semantics that translates such theories into actual pomdps 3. a real time dynamic programming algorithm that produces controllers from such pomdps. We show that the resulting approach is not only clean and general but that is practical as well. We have implemented a shell that accepts pomdp theories and produces controllers, and have tested it over a number of problems. In this paper we present the main elements of the approach and report results for the `omelette problem' where the resulting controller exhibits a better performance than the handcrafted controller. Introduction Consider an agent that has a large supply of eggs and whose goal is to get three good eggs and no bad ones into one of two bowls. The eggs can be either good or bad, and at any...

This work is motivated by the existence of two useful but quite different knowledge representation formalisms, the situation calculus due to McCarthy, and the logic OL of only knowing due to Levesque. In this paper, we propose the logic AOL, which combines both approaches in a clean and natural way. We present a semantics for AOL which generalizes the semantics of OL to account for actions, and a sound and complete set of axioms for AOL which generalizes the Lin and Reiter foundational axioms of the situation calculus to account for only knowing.

The evaluation of an implication by Imaging is a logical technique developed in the framework of modal logic. Its interpretation in the context of a \possible worlds " semantics is very appealing for IR. In 1989, Van Rijsbergen suggested its use for solving one of the fundamental problems of logical models in IR: the evaluation of the implication d! q (where d and q are respectively a document and a query representation). Since then, others have tried to follow that suggestion proposing models and applications, though without much success. Most of these approaches had as their basic assumption the consideration that \a document is a possible world". We propose instead an approach based on a completely di erent assumption: \a term is a possible world". This approach enables the exploitation of term{ term relationships which are estimated using an information theoretic measure. 1 The use of non-classical logic in Information

Amal J. Ahmed Andrew W. Appel # Roberto Virga Princeton University {amal,appel,rvirga}@cs.princeton.edu Abstract We demonstrate a semantic model of general references --- that is, mutable memory cells that may contain values of any (statically-checked) closed type, including other references. Our model is in terms of execution sequences on a von Neumann machine

GOLOG is a high-level programming language for the specification of complex actions. It combines the situation calculus with control structures known from conventional programming languages. Given a suitable axiomatization of what the world is like initially and how the primitive actions change the world, the GOLOG interpreter derives for each program a corresponding linear sequence of legally executable primitive actions, if one exists. Despite its expressive power, GOLOG's applicability is severely limited because the derivation of a linear sequence of actions requires that the outcome of each action is known beforehand. Sensing actions do not meet this requirement since their outcome can only be determined by executing them and not by reasoning about them. In this paper we extend GOLOG by incorporating sensing actions. Instead of producing a linear sequence of actions, the new interpreter yields a tree of actions. The idea is that a particular path in the tree represents a legal execution of primitive actions conditioned on the possible outcome of sensing actions along the way.

The authors outline a theory of conditionals of the form If A then C and If A then possibly C. The 2 sorts of conditional have separate core meanings that refer to sets of possibilities. Knowledge, pragmatics, and semantics can modulate these meanings. Modulation can add information about temporal and other relations between antecedent and consequent. It can also prevent the construction of possibilities to yield 10 distinct sets of possibilities to which conditionals can refer. The mental representation of a conditional normally makes explicit only the possibilities in which its antecedent is true, yielding other possibilities implicitly. Reasoners tend to focus on the explicit possibilities. The theory predicts the major phenomena of understanding and reasoning with conditionals. You reason about conditional relations because much of your knowledge is conditional. If you get caught speeding, then you pay a fine. If you have an operation, then you need time to recuperate. If you have money in the bank, then you can cash a check. Conditional reasoning is a central part of thinking, yet people do not always reason correctly. The lawyer Jan Schlictmann in a celebrated trial (see Harr, 1995, pp. 361–362) elicited the following information from an expert witness about the source of a chemical pollutant trichloroethylene (TCE):

Most agents can acquire information about their environments as they operate. A good plan for such an agent is one that not only achieves the goal, but is also executable, i.e., ensures that the agent has enough information at every step to know what to do next. In this paper, we present a formal account of what it means for an agenttoknow how to execute a plan and to be able to achieve a goal. Such a theory is a prerequisite for producing specifications of planners for agents that can acquire information at run time. It is also essential to account for cooperation among agents. Our account is more general than previous proposals, correctly handles programs containing loops, and incorporates a solution to the frame problem. It can also be used to prove programs containing sensing actions correct.

The knowledge required for action is generally indexical rather than objective. For example, a robot that knows the relative position of an object is generally able to go and pick it up; he need not know its absolute position. Agents may have very incomplete knowledge of their situation in terms of what objective facts hold and still be able to achieve their goals. This paper presents a formal theory of knowledge and action, embodied in a modal logic, that handles the distinction between indexical and objective knowledge and allows a proper specification of the knowledge prerequisites and effects of action. Several kinds of robotics situations involving indexical knowledge are formalized within the framework; these examples show how actions can be specified so as to avoid making excessive requirements upon the knowledge of agents. 1 Introduction 1.1 Motivation Designing autonomous robots or other kinds of agents that interact in sophisticated ways with their environment is hard; yo...