Stochastic Predictive Control under Intermittent Observations and Unreliable Actions

TL;DR

This paper introduces a provably stabilizing control approach for linear systems with intermittent observations and unreliable control transmissions, using Kalman filtering and online quadratic programming.

Contribution

It presents a new method combining Kalman filtering and stochastic optimal control to ensure stability under unreliable communication channels.

Findings

The approach guarantees stability for Lyapunov stable systems.

It handles any positive control bounds and non-zero Bernoulli channel probabilities.

The method is computationally tractable and implementable online.

Abstract

We propose a provably stabilizing and tractable approach for control of constrained linear systems under intermittent observations and unreliable transmissions of control commands. A smart sensor equipped with a Kalman filter is employed for the estimation of the states from incomplete and corrupt measurements, and an estimator at the controller side optimally feeds the intermittently received sensor data to the controller. The remote controller iteratively solves constrained stochastic optimal control problems and transmits the control commands according to a carefully designed transmission protocol through an unreliable channel. We present a (globally) recursively feasible quadratic program, which is solved online to yield a stabilizing controller for Lyapunov stable linear time invariant systems under any positive bound on control values and any non-zero transmission probabilities of Bernoulli channels.