Catálogo de publicaciones - libros

This paper extends a programming language for implementing cognitive agents with the capability to explicitly represent beliefs and reason about them. In this programming language, the beliefs of agents are implemented by modal logic programs, where beliefs are represented by explicit modal operators. A distinction is made between a belief base language that can be used to represent an agent’s beliefs, and a belief query language that can be used to express queries to the agent’s belief base. We adopt and modify a proof procedure that decides if a belief query formula is derivable from the belief base of an agent. We show that the presented proof procedure is sound.

Existing cognitive agent programming languages that are based on the BDI model employ logical representation and reasoning for implementing the beliefs of agents. In these programming languages, the beliefs are assumed to be certain, i.e. an implemented agent can believe a proposition or not. These programming languages fail to capture the underlying uncertainty of the agent’s beliefs which is essential for many real world agent applications. We introduce Dempster-Shafer theory as a convenient method to model uncertainty in agent’s beliefs. We show that the computational complexity of Dempster’s Rule of Combination can be controlled. In particular, the certainty value of a proposition can be deduced in linear time from the beliefs of agents, without having to calculate the combination of Dempster-Shafer mass functions.

An agent who bases his actions upon explicit logical formulae has at any given point in time a finite set of formulae he has computed. Closure or consistency conditions on this set cannot in general be assumed – reasoning takes time and real agents frequently have contradictory beliefs. This paper discusses a formal model of knowledge as explicitly computed sets of formulae. It is assumed that agents represent their knowledge syntactically, and that they can only know finitely many formulae at a given time. In order to express interesting properties of such finite syntactic epistemic states, we extend the standard epistemic language with an operator expressing that an agent knows at most a particular finite set of formulae, and investigate axiomatization of the resulting logic. This syntactic operator has also been studied elsewhere without the assumption about finite epistemic states. A strongly complete logic is impossible, and the main results are non-trivial characterizations of the theories for which we can get completeness. The paper presents a part of a general abstract theory of resource bounded agents. Interesting results, e.g., complex algebraic conditions for completeness, are obtained from very simple assumptions, i.e., epistemic states as arbitrary finite sets and operators for knowing at least and at most.

This paper is concerned with designing architectures for rational agents. In the proposed architecture, agents have belief bases that are theories in a multi-modal, higher-order logic. Belief bases can be modified by a belief acquisition algorithm that includes both symbolic, on-line learning and conventional knowledge base update as special cases. A method of partitioning the state space of the agent in two different ways leads to a Bayesian network and associated influence diagram for selecting actions. The resulting agent architecture exhibits a tight integration between logic, probability, and learning. Two illustrations of the agent architecture are provided, including a user agent that is able to personalise its behaviour according to the user’s interests and preferences.

Multi-agent systems (MAS) can take many forms depending on the characteristics of the agents populating them. Amongst the more demanding properties with respect to the design and implementation of multi-agent system is how these agents may individually reason and communicate about their knowledge and beliefs, with a view to cooperation and collaboration. In this paper, we present a deductive reasoning multi-agent platform using an extension of answer set programming (ASP). We show that it is capable of dealing with the specification and implementation of the system’s architecture, communication and the individual agent’s reasoning capacities. Agents are represented as Ordered Choice Logic Programs (OCLP) as a way of modelling their knowledge and reasoning capacities, with communication between the agents regulated by uni-directional channels transporting information based on their answer sets. In the implementation of our system we combine the extensibility of the JADE framework with the flexibility of the OCT front-end to the Smodels answer set solver. The power of this approach is demonstrated by a multi-agent system reasoning about equilibria of extensive games with perfect information.

We propose a distributed architecture to endow multi-agent systems with a social layer in which normative positions are explicitly represented and managed via rules. Our rules operate on a representation of the states of affairs of a multi-agent system. We define the syntax and semantics of our rules and an interpreter; we achieve greater precision and expressiveness by allowing constraints to be part of our rules. We show how the rules and states come together in a distributed architecture in which a team of administrative agents employ a tuple space to guide the execution of a multi-agent system.

In this paper we cope with providing an approach to declarative semantics of logic-based agent-oriented languages, taking then as a case-study the language DALI which has been previously defined by the authors. This “evolutionary semantics” does not resort to a concept of state: rather, it models reception of events as program transformation steps, that produce a “program evolution” and a corresponding “semantic evolution”. Communication among agents and multi-agent systems is also taken into account. The aim is that of modeling agent’s evolution according to either external (environmental) or internal changes in a logical way, thus allowing in principle the adoption of formal verification methods. We also intend to create a common ground for relating and comparing different approaches/languages.

This paper deals with a goal-oriented agent model called Goal Decomposition Tree (GDT) allowing both to specify and validate the behaviour of an agent. This work takes place in a global framework whose goal is to define a process allowing to start from a problem specification to obtain a validated implementation of a corresponding MAS. The GDT model has been used to specify a prey-predator system which has been verified this way.

Agents need to be able to change their beliefs; in particular, they should be able to contract or remove a certain belief in order to restore consistency to their set of beliefs, and revise their beliefs by incorporating a new belief which may be inconsistent with their previous beliefs. An influential theory of belief change proposed by Alchourron, Gärdenfors and Makinson (AGM) [1] describes postulates which rational belief revision and contraction operations should satisfy. The AGM postulates are usually taken as characterising idealised rational reasoners, and the corresponding belief change operations are considered unsuitable for implementable agents due to their high computational cost [2]. The main result of this paper is to show that an efficient (linear time) belief contraction operation nevertheless satisfies all but one of the AGM postulates for contraction. This contraction operation is defined for an implementable rule-based agent which can be seen as a reasoner in a very weak logic; although the agent’s beliefs are deductively closed with respect to this logic, checking consistency and tracing dependencies between beliefs is not computationally expensive. Finally, we give a non-standard definition of belief revision in terms of contraction for our agent.

Developing applications that make effective use of machine-readable knowledge sources as promised by the Semantic Web vision is attracting much of current research interest; this vision is also affecting important trends in computer science such as grid-based and ubiquitous computing. In this paper, we formally define a version of the BDI agent-oriented programming language AgentSpeak based on description logic rather than predicate logic. In this approach, the belief base of an agent contains the definition of complex concepts, besides specific factual knowledge. We illustrate the approach using examples based on the well-known smart meeting-room scenario. The advantages of combining AgentSpeak with description logics are: (i) queries to the belief base are more expressive as their results do not rely only on explicit knowledge but can be inferred from the ontology; (ii) the notion of belief update is refined given that (ontological) consistency of a belief addition can be checked; (iii) retrieving a plan for handling an event is more flexible as it is not based solely on unification but on the subsumption relation between concepts; and (iv) agents may share knowledge by using ontology languages such as OWL. Extending agent programming languages with description logics can have a significant impact on the development of multi-agent systems for the semantic web.