Total variations reduction for an exact control applied to a dynamical mechanical system

TL;DR

This paper introduces a smoothing control strategy for dynamical mechanical systems that minimizes energy expenditure and oscillations by incorporating total variation reduction into the optimal control criterion, using asymptotic methods.

Contribution

It develops an exact control approach for non-differentiable cost criteria that reduces control variations and energy use, based on asymptotic analysis and Tykhonov regularization.

Findings

Control variations are effectively reduced with parameter tuning.

The method minimizes control energy among exact controls.

Validated on three example systems.

Abstract

In an optimal control strategy, an important point is to define the cost of the control. Usually it is added to the control criterion and multiplied by a small coefficient denoted by $\varepsilon$ which is known as the marginal cost of the control. The key idea of this paper, is to introduce a smoothing term in the control cost which aims at reducing the quantity of energy spent and reducing the oscillations of the control. Then using a so-called asymptotic control based on the smallness of $\varepsilon$, we construct an exact control which can be implemented in a close loop. The energy involved in the control depends mainly on the variation of the control. Therefore it seems natural to include this quantity (the total variations) in the criterion involved in the optimal control. This can be done approximately by introducing the $L^1$ norm of the first order derivative of the control. The control strategy that we develop in this paper can be applied to such linear models. One important and new point is that we focus on exact control strategies for a non-differentiable criterion because of the cost of the control. Following the ideas of Tykhonov regularization method, it is proved using the so-called asymptotic method based on the smallness of the marginal cost of the control, that the exact control suggested is the one which represents the minimum of the marginal cost among exact controls. Furthermore, and it is the main technical point, it can reduce the variations of the control with an adequate tuning of the various parameters of the control loop. We test the method on three examples.