Jean-Christophe Routier

Halting Problem of One Binary Horn Clause is Undecidable

This paper proposes a codification of the halting problem of any Turing machine in the form of only one right-linear binary Horn clause as follows: p(t) ← p(tt). where t (resp. tt) is any (resp. linear) term. Recursivity is well-known to be a crucial and fundamental concept in programming theory. This result proves that in Horn clause languages there is no hope to control it without additional hypotheses even for the simplest recursive schemes.

Smallest horn clause programs

Abstract The simplest nontrivial program pattern in logic programming is the following where fact, goal, left , and right are arbitrary terms. Because the well-known append program matches this pattern, we will denote such programs “ append -like.” In spite of their simple appearance, we prove in this paper that termination and satisfiability (i.e., the existence of answer-substitutions, called the emptiness problem) for append -like programs are undecidable. We also study some subcases depending on the number of occurrences of variables in fact, goal, left , or right . Moreover, we prove that the computational power of append -like programs is equivalent to the one of Turing machines; we show that there exists an append -like universal program. Thus, we propose an equivalent of the Bohm-Jacopini theorem for logic programming. This result confirms the expressiveness of logic programming. The proofs are based on program transformations and encoding of problems, unpredictable iterations within number theory defined by J. H. Conway, or the Post correspondence problem.

A Virtual Selling Agent which is Proactive and Adaptive

In this paper, we claim that the online selling process can be improved if the experience of the customer is closer to the one in a retailing store. For this purpose, we aim at providing a virtual selling agent that is proactive and adaptive. Our proactive dialogical agent initiates the dialogue, uses marketing strategies and drives the inquiring process for collecting information in order to make relevant proposals. Moreover, our virtual seller is adaptive since she is able to adjust her behaviour according to the buyer profile.

RIO: roles, interactions and organizations

The notions of role and organization have often been emphasized in several agent oriented methodologies. Sadly, the notion of interaction has seldom been reified in these methodologies. We define here a model of runnable specification of interaction protocols. Then, we propose a methodology for the design of open multi-agent systems based on an engineering of interaction protocols. These interaction protocols are described in term of conversation between micro-roles characterized by their skills, then micro-roles are gathered in composite roles. Then, composite roles are used to build abstract agents. Lastly, these latter can be distributed on running agents of a multi-agent system.

Bilateral Negotiation of a Meeting Point in a Maze

Negotiation between agents aims at reaching an agreement in which the conflicting interests of agents are accommodated. In this paper, we present a concrete negotiation scenario where two agents are situated in a maze and the negotiation outcome is a cell where they will meet. Based on their individual preferences (a minimal distance from their location computed from their partial knowledge of the environment), we propose a negotiation protocol which allows agents to submit more than two proposals at the same time and a conciliatory strategy. Formally, we prove that the agreement reached by such a negotiation process is Pareto- optimal and a compromise, i.e. a solution which minimizes the maximum effort for one agent. Moreover, the path between the two agents emerges from the repeated negotiations in our experiments.

A Multi-Agent Approach to Co-operative Work

This paper presents a multi-agent approach for user interface design of co-operative applications. We show here that this approach provides advantages from multi-agent researches about intelligence and modeling. It allows the designer to develop more easily the user interface of his application. Two different aspects are improved: first, the management of the interactions between components of the interface and second, the interactions between the user and the application, in particular thanks to intelligent assistants. Moreover, an agent oriented analysis of the problem highlights the different role and skills and thus eases the programmer task and the evolution capacity of the application. To illustrate our purpose, we present a co-operative work application built with our multi-agent development framework, called Magique. The main goal of this article is not really to present a groupware application but rather to show how it can be easily developed, maintained and extended thanks to Magique and its multi-agent approach

Planning for spatially situated agents

Most often, agent-based situated simulations use reactive agents. While the use of this paradigm offers a rather natural way to build simple behaviours, it is not so easy to define complex behaviours, and these are often built ad hoc and requires precise adjustments. Moreover, reactive model is not well fitted to describe deliberate group behaviours since it is more intended to emergence. For these reasons and in order to obtain more explicative models, we are working on proactive agent-based situated simulations. We aim at proposing a generic cognitive agent model. Such agents must build plan to achieve some goals and thus to propose a behaviour. We show why and how classical planning works must be adapted to the particular context of spatially situated agents in simulations.

Towards a Pragmatic Methodology for Open Multi-agent Systems

This paper introduces the Rio methodology, which relies on the notions of Role, Interaction and Organization. We define a model of runnable specification of interaction protocols that represents interactions between agents globally, and that can be processed in orsder to generate Colored Petri Nets that handle agent conversations. Moreover, these specifications are not tied to a particular agent model and could therefore be used to enable multi-agent systems interoperability.

Behavior Design of Game Character's Agent

In human-level simulations, like video games can be, the design of characters behaviors has an important impact on simulation realism. We propose to divide it into a reasoning part, dedicated to a planner, and an individuality part, assigned to an action selection mechanism. Applying the separation of declarative and procedural aspects, the principle is to provide every characters agent with the same procedural mechanisms: the planner and the action selection mechanism. Declarative knowledge is then used at the agent level to individualize the behavior. The contribution of this paper consists in a motivation based action selection mechanism that allows individualization in behavior. The modularity provided by the motivations enables a large variety of behaviors for which the designer has to choose parameters. If the simulation of characters are our first motivation, the principles involved in the proposed motivation based action mechanism are general enough to be used in other contexts.In human-level simulations, like video games can be, the design of characters behaviors has an important impact on simulation realism. We propose to divide it into a reasoning part, dedicated to a planner, and an individuality part, assigned to an action selection mechanism. Applying the separation of declarative and procedural aspects, the principle is to provide every characters agent with the same procedural mechanisms: the planner and the action selection mechanism. Declarative knowledge is then used at the agent level to individualize the behavior. The contribution of this paper consists in a motivation based action selection mechanism that allows individualization in behavior. The modularity provided by the motivations enables a large variety of behaviors for which the designer has to choose parameters. If the simulation of characters are our first motivation, the principles involved in the proposed motivation based action mechanism are general enough to be used in other contexts.

Towards a pragmatic use of ontologies in multi-agent platforms

The knowledge representation field is gaining momentum with the advent of the Semantic Web activity within the W3C. This working group, thanks to previous researches, has proposed the Ontology Web Language to enhance the expressivity of web pages and to allow semantic inferences. This paper argues that knowledge representation technologies should be core components of multi-agent platforms.