Distributed backtracking algorithm based on tree decomposition over wireless sensor networks
Résumé
In this paper, we propose a methodological approach to solve distributed nonbinary constraint satisfaction problem (CSP) on wireless sensor networks (WSNs). A distributed CSP is a CSP in which variables and constraints are distributed among multiple agents. On WSNs, it is usual to handle applications that need to solve distributed problems. Different real-world applications can be modeled as distributed CSPs, and numerous algorithms based on enumerative search have been proposed to solve them. The most cited one is distributed backtracking algorithm in which each variable is associated to each agent. This algorithm is known as fine-grained distributed algorithm. All the search efforts of this algorithm concerns the communication between agents that are very expensive. In addition, this approach is not realistic because, in general, an agent might control more than one variable. In this paper, we propose a generic methodology for developing coarse-grained backtracking algorithm. Mainly, a preprocess technique breaks a single large problem into a set of smaller connected ones. These semi-independent CSPs can be efficiently and concurrently solved and can cooperate to solve the whole problem. We illustrate the preprocess technique by the tree decomposition method for its good theoretical properties. The aim of our paper is to present an efficient approach to solve complex distributed CSPs over WSNs. Copyright (c) 2011 John Wiley & Sons, Ltd.