-
Online Optimization and Learning in Uncertain Dynamical Environments with Performance Guarantees
Authors:
Dan Li,
Dariush Fooladivanda,
Sonia Martinez
Abstract:
We propose a new framework to solve online optimization and learning problems in unknown and uncertain dynamical environments. This framework enables us to simultaneously learn the uncertain dynamical environment while making online decisions in a quantifiably robust manner. The main technical approach relies on the theory of distributional robust optimization that leverages adaptive probabilistic…
▽ More
We propose a new framework to solve online optimization and learning problems in unknown and uncertain dynamical environments. This framework enables us to simultaneously learn the uncertain dynamical environment while making online decisions in a quantifiably robust manner. The main technical approach relies on the theory of distributional robust optimization that leverages adaptive probabilistic ambiguity sets. However, as defined, the ambiguity set usually leads to online intractable problems, and the first part of our work is directed to find reformulations in the form of online convex problems for two sub-classes of objective functions. To solve the resulting problems in the proposed framework, we further introduce an online version of the Nesterov accelerated-gradient algorithm. We determine how the proposed solution system achieves a probabilistic regret bound under certain conditions. Two applications illustrate the applicability of the proposed framework.
△ Less
Submitted 17 February, 2021;
originally announced February 2021.
-
Data-driven Predictive Control for a Class of Uncertain Control-Affine Systems
Authors:
Dan Li,
Dariush Fooladivanda,
Sonia Martinez
Abstract:
This paper studies a data-driven predictive control for a class of control-affine systems which is subject to uncertainty. With the accessibility to finite sample measurements of the uncertain variables, we aim to find controls which are feasible and provide superior performance guarantees with high probability. This results into the formulation of a stochastic optimization problem (P), which is i…
▽ More
This paper studies a data-driven predictive control for a class of control-affine systems which is subject to uncertainty. With the accessibility to finite sample measurements of the uncertain variables, we aim to find controls which are feasible and provide superior performance guarantees with high probability. This results into the formulation of a stochastic optimization problem (P), which is intractable due to the unknown distribution of the uncertainty variables. By develo** a distributionally robust optimization framework, we present an equivalent and yet tractable reformulation of (P). Further, we propose an efficient algorithm that provides online suboptimal data-driven solutions and guarantees performance with high probability. To illustrate the effectiveness of the proposed approach, we consider a highway speed-limit control problem. We then develop a set of data-driven speed controls that allow us to prevent traffic congestion with high probability. Finally, we employ the resulting control method on a traffic simulator to illustrate the effectiveness of this approach numerically.
△ Less
Submitted 29 April, 2021; v1 submitted 22 November, 2019;
originally announced November 2019.
-
Data-driven Variable Speed Limit Design for Highways via Distributionally Robust Optimization
Authors:
Dan Li,
Dariush Fooladivanda,
Sonia Martinez
Abstract:
This paper introduces an optimization problem (P) and a solution strategy to design variable-speed-limit controls for a highway that is subject to traffic congestion and uncertain vehicle arrival and departure. By employing a finite data-set of samples of the uncertain variables, we aim to find a data-driven solution that has a guaranteed out-of-sample performance. In principle, such formulation l…
▽ More
This paper introduces an optimization problem (P) and a solution strategy to design variable-speed-limit controls for a highway that is subject to traffic congestion and uncertain vehicle arrival and departure. By employing a finite data-set of samples of the uncertain variables, we aim to find a data-driven solution that has a guaranteed out-of-sample performance. In principle, such formulation leads to an intractable problem (P) as the distribution of the uncertainty variable is unknown. By adopting a distributionally robust optimization approach, this work presents a tractable reformulation of (P) and an efficient algorithm that provides a suboptimal solution that retains the out-of-sample performance guarantee. A simulation illustrates the effectiveness of this method.
△ Less
Submitted 26 October, 2018;
originally announced October 2018.
