Estimation and control applications of linear consensus algorithms
Abstract:
In this talk I will present some applications in the context of estimation and control of a popular class of linear distributed algorithms, known as consensus algorithms, which are particularly suitable for computing averages of local quantities. These applications include sensor calibration, least square parameter estimation, event detection, and clock synchronization. I will also show that these applications give rise to performance metrics which are different from the essential spectral radius of the consensus matrix, which is the commonly used performance index. More precisely, these metrics are typically convex combinations of all matrix eigenvalues. Finally, I will briefly show that this fact has important consequences in term of optimal design of the consensus matrix.