Performance of Distributed Sensing Algorithms with Correlated Noise and Defective Sensors

Abstract

Correlated noise is studied in the context of distributed sensing problems. Previous work has examined the problem of performing distributed estimation in the presence of sensor defects. These defects disrupt sensor measurements but allow the other subsystems of the sensor node to operate normally. Provably optimal and fully distributed estimation algorithms exist for this scenario. A series of computer simulation experiments was conducted to measure how these estimation algorithms are affected by the presence of correlations between various random quantities in the problem setup. Numerical results demonstrate that, in particular network configurations, correlations between sensor defects and communication links can practically alter the performance of distributed estimation algorithms.