Design of unbiased state estimators for WSNs with consensus on measurements and estimates and improved robustness

Wireless sensor networks (WSNs) is a technology with important developments in recent years. Its incursion in areas such as healthcare, industry and services has been steadily increasing, mainly due to the miniaturization of electronics and the growing acceptance of cyber-physical systems. However, a very important subject of research continues to be the development of estimators with the robustness needed for the harsh conditions associated with the WSNs applications. Moreover, such estimators should comply with the unique characteristics imposed by the WSNs like scalability, energy saving and redundancy, while maintaining a consensus on the network. A very popular algorithm for optimal estimation is the Kalman filter (KF). Many works have implemented it as a sensor fusion technique in WSNs, due to its optimality. However it cannot guarantee the robustness needed in real life implementations. In this work we develop a set of robust estimators based on unbiased finite impulse response (UFIR) filters to address the lack of robustness of the popular KF. The developed filters are adequate to be implemented in WSNs. The algorithms have been tested against similar filters based on KF with simulated and real data, showing better results in terms of estimation error reduction, where the smallest improvement was of 1.4 percent in terms of the root mean squared error. We even produce accurate results in applications where KF could not be implemented. The developed filters attained better robustness against model errors, unknown statistics, and missing measurements.