Optimal Estimation with Limited Measurements and Noisy Communication

TL;DR

This paper investigates optimal sensor scheduling and encoding strategies for sequential state estimation over noisy channels with limited transmissions, revealing surprising phase-transition phenomena in transmission usage.

Contribution

It introduces the first analysis of optimal encoding and scheduling policies in a limited communication setting with noisy channels, extending prior noiseless models.

Findings

Optimal policies derived for specific source and noise densities.

Numerical analysis shows phase-transition in transmission opportunities.

Noiseless and noisy communication models exhibit different behaviors.

Abstract

This paper considers a sequential estimation and sensor scheduling problem with one sensor and one estimator. The sensor makes sequential observations about the state of an underlying memoryless stochastic process, and makes a decision as to whether or not to send this measurement to the estimator. The sensor and the estimator have the common objective of minimizing expected distortion in the estimation of the state of the process, over a finite time horizon, with the constraint that the sensor can transmit its observation only a limited number of times. As opposed to the prior work where communication between the sensor and the estimator was assumed to be perfect (noiseless), in this work an additive noise channel with fixed power constraint is considered; hence, the sensor has to encode its message before transmission. For some specific source and channel noise densities, we obtain the optimal encoding and estimation policies in conjunction with the optimal transmission schedule. The impact of the presence of a noisy channel is analyzed numerically based on dynamic programming. This analysis yields some rather surprising results such as a phase-transition phenomenon in the number of used transmission opportunities, which was not encountered in the noiseless communication setting.