Routing for reliability in decentralised sensing networks

Considers the problem of management and configuration in a decentralised sensing network in which each sensor node implements both local and global estimation algorithms. The sensing networks considered here are characterised by three general constraints. The authors are concerned with the problem of management in such systems: they require that the network be able to overcome local failures by supporting topology reconfiguration and the reallocation of sensing and communication tasks. Previous work has proposed that this can be achieved only by relaxing the system constraints. The authors show that it is necessary to relax the system constraints only in a very limited sense. The approach taken allows the authors to maintain the decentralised system constraints on the sensing nodes. A fundamental aspect of management is the provision of a means of routing information which is not compromised by failures of communication links or processors and which, in turn, does not compromise the general constraints under which the decentralised system and estimation algorithms are required to operate. The authors provide a reliable routing system, using an algorithm for asynchronous distributed dynamic programming and describe its implementation on a practical decentralised sensing network for a process monitoring plant.