Self-reflective model predictive control

TL;DR

This paper introduces a self-reflective model predictive control scheme that accounts for its own limitations due to noise and measurement errors by propagating both forward and backward matrices to optimize control performance.

Contribution

It presents a novel MPC controller that propagates an adjoint state backward in time to explicitly consider its own limitations, unlike existing methods.

Findings

Demonstrated effectiveness on a case study

Accounts for process noise and measurement errors

Improves control performance by self-assessment

Abstract

This paper proposes a novel control scheme, named self-reflective model predictive control, which takes its own limitations in the presence of process noise and measurement errors into account. In contrast to existing output-feedback MPC and persistently exciting MPC controllers, the proposed self-reflective MPC controller does not only propagate a matrix-valued state forward in time in order to predict the variance of future state-estimates, but it also propagates a matrix-valued adjoint state backward in time. This adjoint state is used by the controller to compute and minimize a second order approximation of its own expected loss of control performance in the presence of random process noise and inexact state estimates. The properties of the proposed controller are illustrated with a small but non-trivial case study.