Linear Sensor Collaboration for Distributed Parameter Estimation in the Presence of Communication Failures

The problem of scalar parameter estimation in a distributed wireless sensor network (WSN), in the presence of communication failures, is considered in this work. When sensors obtain measurements and attempt to transmit their measurements to the fusion center (FC) for parameter estimation, the transmissions to the FC may be unsuccessful due to various reasons such as poor communication channels, large distance between the sensors and FC or insufficient transmit power. To overcome the degradation of estimation performance due to missing data, we consider linear inter-sensor collaboration, where sensors exchange measurements with neighboring sensors, before transmitting to the FC. We consider two objectives: 1) maximize the estimation accuracy subject to collaboration power constraints and 2) minimize the collaboration power subject to the required estimation accuracy. We consider linear estimators for the parameter inference task, and propose methods for designing the collaboration scheme (collaboration weights). The performances of the estimators and collaboration design are compared using numerical results and simulations.