A Diffusion-Based EM Algorithm for Distributed Estimation in Unreliable Sensor Networks

Distributed parameter estimation in unreliable sensor networks via broadcast gossip algorithms

“…Recently, the paper [1] proposes a distributed EM algorithm to be implemented in a WSN with a decentralized architecture using diffusion-like strategies. In [2], the authors design a consensus-based algorithm for distributed estimation problem with linear or nonlinear measurements.…”

“…Further improvement of this work is presented in [3] to handle the case: WSNs with randomly switching topologies. Of note is that all of them [1]- [3] are developed to deal with the unreliable environment including only link failures based on synchronous updates framework. Though synchronous algorithms have advantages in convergence time, they accumulate more and more quantization and measurement noise over time, which greatly affect the computations precision.…”

“…First, our algorithm is fully asynchronous thus can effectively reduce the accumulated measurement and quantization noise, and further enhance the computation precision than algorithms based on synchronous updates [1]- [3]. In addition, our algorithm does not require error recovery mechanisms thus can solve gracefully distributed parameter estimation over networks with a huge number of nodes.…”

“…Furthermore, we wish to obtain a quantized message q[x ij ] with length ℓ bits, where ℓ represents the quantization precision. Therefore, we have L = 2 ℓ quantization points given by the set Π = {π 1 …”

“…In practice, WSNs may be unreliable [1]- [3] due to external malicious attacks, energy exhausting, incorrect sensing and other factors. A convenient way when dealing with agreement in unreliable networks is to exchange messages by an asynchronous framework.…”