… controller with inaccurate predictions, obtained here in terms of the prediction error norm. …
on the control of linear timeinvariant (LTI) systems subject to process noise. The optimal control …

… results on regret minimization for the linear quadratic regulator to optimal control subject to
… allows to reduce a regretsuboptimal control problem to a suboptimal H∞ problem might be …

… a criterion for controller design we focus on regret, defined as … is H∞ control [1], for bounded
amplitude it is l1 control). In this paper, we consider optimal control through the lens of regret …

… Motivated by learning theory, as a criterion for controller design we focus on regret, defined
… perspective to expose the fundamental difference between H∞ and regretoptimal controls. …

… of our controller follows from our reduction to H∞ control; … on minimizing regret against a
fixed controller from some parametric class of control policies (policy regret), while we focus on …

… H∞ Estimation and Control Our results rely heavily on techniques from robust estimation and
control… of finding a regret-optimal estimator can be reduced to an H∞ estimation problem. …

… of regret minimization. Unlike most prior work in this area, we focus on the problem of designing
an online controller … cost, whereas in H∞ control w and v are assumed to be generated …

… h} is the observed sequence of controls. A desirable … is to be no-regret, ie the time averaged
version of the regret … regret bound whose leading term regret depends linearly on J⋆. …

… ellipsoidal bounds on the disturbance, we show that the dynamic regret bound can be …
Finally, we show that the proposed method can outperform the H2- and H∞ controllers on a mass-…

… Thus, we will focus on the dynamic regret analysis and study how predictions can
improve the dynamic regret. Example 1 (Linear quadratic (LQ) tracking). Consider a discrete-time …