CCP07-IPSN
From VaVMURI
This paper presents a unified methodology for designing sensor networks that are analog computing systems, time- stepped synchronous systems, shared-memory asynchronous systems, and reliable or unreliable message-passing systems. The dynamics of the systems may by specified by differential equations, difference equations or temporal logic. Commu- nication and computational resources in the network vary with time due to environmental factors such as noise inter- ference or attacks by adversaries. Computations proceed by opportunistically employing the resources available at each point, progressing rapidly when more resources are avail- able and slowing down when resources become unavailable. The paper proposes mathematical models for a variety of sensor network types and suggests a uniform framework for designing algorithms for them. The impact of the persis- tence of network connectivity on the design of algorithms is explored: if network connectivity remains unchanged for long periods then algorithms that compute global parame- ters of the network are efficient. If, by contrast, the network changes frequently then estimates of global parameters are likely to be incorrect and atomic operations across the entire network are likely to be aborted; in this case algorithms in which agents use local information are more efficient. The paper develops design techniques and performance analysis of systems in which agents use information with different de- grees of locality ranging from immediate neighbors to global parameters.
- Conference submission (restricted)
