Figure 6 - uploaded by Ismel Brito
Content may be subject to copyright.
Source publication
In the last years, the AI community has shown an increasing interest in distributed problem solving. In the scope of distributed constraint reasoning, several asynchronous backtracking procedures have been proposed for finding solutions in a constraint network distributed among several computers. They differ in the way they store failing combinatio...
Context in source publication
Context 1
... 2 Suppose we have a network with 11 agents, where x 1 , . . . , x 10 are all linked to x 11 with an inequality constraint. (See Fig. 6.) x 1 , x 2 , x 3 are assigned value a and send Inf o messages, x 4 , x 5 , x 6 are assigned value b and send Inf o messages, x 7 , x 8 , x 9 are assigned value c and send Inf o messages, and finally, x 10 is assigned value d and sends Inf o message. The domain of x 11 was {a, b, c, d}. Once x 11 has received the Inf o message sent by ...
Citations
... Dynamic Backtracking(DiBT) Cette approche a été proposée au départ par Hamadi et al. mais n'était pas complète [66]. Puis elle a été étendue par plusieurs auteurs [9,12]. C'est un algorithme calqué sur le modèle du Backtrack dynamique distribué. ...
Une large gamme de problèmes pratiques nécessite une diversité de représentation et de modélisation des données et de développer des modèles dans lesquels les différentes données peuvent être représentées. Dans cette thèse, nous nous intéressons à l'hybridation de deux modèles : le modèle des Systèmes Experts (SE) et le modèle des Problèmes de Satisfaction de Contraintes (CSP). L'objectif de notre travail est de proposer un système distribué et sécurisé pour intégrer des contraintes dans un moteur d'inférence. Pour ce faire, nous avons d'une part développé un outil de communication qui facilite la collaboration entre SE et CSP. D'autre part, nous avons conçu un algorithme qui permet la communication entre plusieurs agents situés dans un environnement distribué. Enfin, un de nos buts, et non des moindres, est d'assurer la protection des données privées de chaque entité. La thèse est donc constituée de trois axes principaux : Le premier axe vise l'élaboration d'une méthode de communication entre les deux modèles. Tout d'abord, nous décrivons une procédure de transformation automatique entre un système expert à base de règles vers un nouveau modèle de CSP dynamique nommé DDCSP (Dynamic Domain CSP) que nous avons au préalable défini. Cette procédure de transformation automatique permettra l'injection des résultats de l'un des deux modèles en entrée de l'autre. Cette procédure joue donc un rôle essentiel pour assurer la collaboration qui s'appuie sur l'échange d'informations. Le deuxième axe est consacré à la prise en compte de la distribution d'un problème CSP sur plusieurs sites. Nous proposons un algorithme basé sur la notion de coopération et de parallélisme qui assure une résolution distribuée entre plusieurs agents. Notre approche consiste à construire un anneau d'agents autonomes, responsable chacun d'une partie des variables et des contraintes du problème. Chacun de ces agents va initier un processus qui explore une branche différente de l'arbre de recherche. Des heuristiques sont proposées pour garantir une diversification des explorations, en d'autres termes pour éviter que les branches explorées ne se recouvrent. Enfin, nous présentons une technique de sécurisation de cet algorithme dans l'environnement distribué basée sur l'utilisation judicieuse des propriétés de cryptographie asymétrique pour préserver la confidentialité des instanciations. Afin d'effectuer une validation expérimentale de nos travaux, une implémentation dans les langages de programmation C/C++ ou Java est décrite dans chacun de ces trois axes.
... In general, the messages contain information about assignments of values to variables and rebuttals trust by employees who have no purpose compatible with their own variables. Mainly we mention the Asynchronous Backtracking (ABT) algorithm that was proposed by М. Yokoo [17] and some of its alternatives [18, 19, 20]. These approaches are designed mainly for the treatment of non-binary constraints, however most systems of real constraints are non-binary. ...
In this paper we describe an original computational model for solving different types of Distributed Constraint Satisfaction Problems (DCSP). The proposed model is called Controller-Agents for Constraints Solving (CACS). This model is intended to be used which is an emerged field from the integration between two paradigms of different nature: Multi-Agent Systems (MAS) and the Constraint Satisfaction Problem paradigm (CSP) where all constraints are treated in central manner as a black-box. This model allows grouping constraints to form a subset that will be treated together as a local problem inside the controller. Using this model allows also handling non-binary constraints easily and directly so that no translating of constraints into binary ones is needed. This paper presents the implementation outlines of a prototype of DCSP solver, its usage methodology and overview of the CACS application for timetabling problems.
... In (Silaghi et al., 2001 ) the authors show how an algorithm for maintaining consistency during distributed asynchronous search can be designed by expanding on ABT. In Bessì ere et al., 2005; Bessiére et al., 2003 ) they propose an asynchronous backtracking algorithm ABT not which does not need to add links between initially unconnected agents. However, none of these notices the redundancy problem in the original ABT. ...
... However, none of these notices the redundancy problem in the original ABT. The algorithms in Bessì ere et al., 2005; Bessiére et al., 2003) come close—the authors notice that an ABT algorithm without adding links between initially unconnected agents will also works correctly. However, they did not mention that the links were unnecessary and did not give a proof of this. ...
This paper shows how the Asynchronous Backtracking (Yokoo et al., 1998) algorithm, a well known distributed constraint satisfaction al- gorithm, produces unnecessary messages and introduces our optimized algo- rithm, Message Management Asynchronous Backtracking, which reduces the number of messages the agents send. The message management mechanism removes the redundant messages, keeps message queue updated, and handles messages by package instead of individually in order to improve eciency. Our test results show the algorithm significantly reduces the total number of messages sent and drastically reduces the number of cycles used when solving instances of the graph coloring problem.
... In [6] the authors show how an algorithm for maintaining consistency during distributed asynchronous search can be designed by expanding on ABT. In [2,1] they propose an asynchronous backtracking algorithm ABT not which does not need to add links between initially unconnected agents. We found no paper that notices the redundancy problem in the original ABT. ...
... We found no paper that notices the redundancy problem in the original ABT. The algorithms in [2,1] come close—the authors notice that an ABT algorithm without adding links between initially unconnected agents will also works correctly. However, they did not mention that the links were unnecessary and did not give a proof of this. ...
We show how the Asynchronous Backtracking Algorithm, a well known distributed constraint satisfaction algorithm, produces unnecessary messages. Our new optimized algo-rithm reduces the number of messages by implementing message management mechanism. Tests show our algorithm significantly reduces the total number of messages sent and drasti-cally reduces the number of cycles used when solving instances of the graph coloring problem.
Hidden Markov Models (HMMs) have been successfully applied to problems from different areas over the last years, like biomedical signal processing area. Independent of the application it is well known that an important step considering the use of HMMs is the initialisation of the parameters of the model. This initialisation should take into account the knowledge about the addressed problem and also optimisation techniques to estimate the best initial parameters given a cost function, and consequently, to estimate the best log-likelihood. This paper proposes the initialisation of Hidden Markov Models parameters using the optimisation algorithm Differential Evolution with the aim to obtain the best log-likelihood.
Following the pioneer work of Yokoo and colleagues on the ABT (asynchronous backtracking) algorithm, several ABT-based procedures have been proposed for solving distributed constraint networks. They differ in the way they store nogoods, but they all use additional communication links between unconnected agents to detect obsolete information. In this paper, we propose a new asynchronous backtracking algorithm which does not need to add links between initially unconnected agents. To make the description simpler and to facilitate the comparisons between algorithms, we present a unifying framework from which the new algorithm we propose, as well as existing ones, are derived. We provide an experimental evaluation of these algorithms.
We aim in this paper to present a model called controller-agents for constraints solving or CACS for short. This model is intended to be used for solving distributed constraint satisfaction problem (DCSP) which is an emerged field from the integration between two paradigms of different nature: multi-agent systems (MAS) that is characterized by the autonomy and the distribution of its entities and the constraint satisfaction problem paradigm (CSP) where all constraints are treated in central manner as a black-box. CACS is based on special kind of agents called controllers. A controller role is to encapsulate and verify some constraints assigned to it. This model allows grouping constraints to form a subset that will be treated together as a local problem inside the controller. Using this model allows also handling non-binary constraints easily and directly so that no translating of constraints into binary ones is needed. Based on CACS, a prototype of DCSP solver is built. This paper presents the implementation outlines of that prototype and its usage methodology. The prototype is built in Java using general interfaces of both MAS and CSP platforms. These interfaces allow users to use the platforms of their choice providing that they implement these interfaces with the chosen platforms.
Differential Evolution (DE) is a tool for efficient optimisation, and it belongs to the class of evolutionary algorithms, which include Evolution Strategies and Genetic Algorithms. DE algorithms work well when the population covers the entire search space, and they have shown to be effective on a large range of classical optimisation problems. However, an undesirable behaviour was detected when all the members of the population are in a basin of attraction of a local optimum (local minimum or local maximum), because in this situation the population cannot escape from it. This paper proposes a modification of the standard mechanisms in DE algorithm in order to change the exploration vs. exploitation balance to improve its behaviour. 1
