Article
Engineering, Industrial
Shing Chih Tsai, Jun Luo, Guangxin Jiang, Wei Cheng Yeh
Summary: This article introduces an adaptive fully sequential Ranking-and-Selection (R&S) procedure, adopting the classic Indifference-Zone (IZ) formulation in the statistical literature and incorporating control variates method. The proposed procedure is demonstrated to have advantages through simulation experiments.
Article
Mathematics, Applied
Raghu Pasupathy, Yongjia Song
Summary: This study introduces an adaptive sequential SAA algorithm to solve large-scale two-stage stochastic linear programs, achieving favorable performance through a sequential framework with optimal sample size schedule and the use of warm starts. Extensive numerical tests demonstrate the success of the proposed algorithm, providing a solution with a probabilistic guarantee on quality.
SIAM JOURNAL ON OPTIMIZATION
(2021)
Article
Management
Zhenxia Cheng, Jun Luo, Ruijing Wu
Summary: We study the simulation optimization problem of selecting the best system design, known as ranking and selection (R&S). We propose fully sequential procedures that incorporate adaptive sampling rules while preserving finite-sample statistical guarantees. Specifically, we introduce an adaptive sampling rule that utilizes consecutively updated sample mean and variance information by solving a minimization problem of the approximated total sample size. Extensive simulation experiments demonstrate the efficiency of the proposed procedures, and we apply them to solve an ambulance dispatching problem.
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH
(2023)
Article
Automation & Control Systems
Fengzhong Li, Yungang Liu
Summary: This article discusses the issue of control-dependent stochastic noise in adaptive control. It proposes basic theorems on stochastic convergence and establishes a martingale-based analysis pattern for adaptive control. Global stabilization of a certain class of uncertain nonlinear systems is achieved through the use of dynamic gains.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2023)
Article
Operations Research & Management Science
Jan Seidler, Ondrej Tybl
Summary: We study a continuous-time Robbins-Monro-type stochastic approximation procedure for a system described by a stochastic differential equation driven by a general Levy process, and we establish sufficient conditions for its convergence using Lyapunov functions. Despite the possible disruption caused by the jump part of the noise, we show that convergence can still be achieved by choosing suitable noise coefficients, even under weaker assumptions on the drift compared to the diffusion case or in the presence of multiple roots of the drift.
JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Tandong Li, Shaobo Li, Junxing Zhang, Hang Sun, Chaojie Zheng, Dongchao Lv
Summary: A new hybrid controller is proposed for high-performance tracking control of permanent magnet synchronous motors in perturbed environments. The controller achieves full-state performance constraints using a prescribed performance method and avoids complexity explosion using a time-varying filter. By combining Lyapunov-Krasovskii functionals with adaptive neural networks, the controller solves the problems of time-delay disturbance and unknown nonlinear dynamics.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2023)
Article
Automation & Control Systems
Hongjiang Qian, George Yin, Qing Zhang
Summary: This article presents a new deep learning framework for general nonlinear filtering. The main contribution is the development of a computationally feasible procedure. The proposed algorithms can handle challenging filtering problems involving diffusions with randomly-varying switching. The article demonstrates the efficiency of the algorithm through highly nonlinear dynamic system examples.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2023)
Article
Automation & Control Systems
Vasileios Tzoumas, Ali Jadbabaie, George J. Pappas
Summary: In this article, the authors propose a robust and adaptive maximization algorithm for solving discrete optimization problems in adversarial environments. The algorithm, called RAM, runs in an online fashion and adapts to the history of failures in each step. It guarantees near-optimal performance and has both provable per-instance a priori bounds and tight and/or optimal a posteriori bounds.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2022)
Article
Automation & Control Systems
Wuquan Li, Miroslav Krstic
Summary: A new least-squares identification scheme is proposed for stochastic strict-feedback nonlinear systems with unknown parameters, without regressor filtering. The key new element in the estimator design is a weighted term with design parameters, introduced to handle nonlinear terms and stochastic noise. Adaptive controllers are designed to ensure global stability in probability at the equilibrium point and regulation of states to zero almost surely.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2021)
Article
Automation & Control Systems
Zhaorong Zhang, Juanjuan Xu, Xun Li
Summary: This paper investigates a discrete-time stochastic control problem with linear quadratic criteria over an infinite-time horizon. The focus is on control systems whose system matrices are associated with random parameters involving unknown statistical properties. A distributed stochastic approximation algorithm is designed to solve the Riccati equation and obtain the optimal controller for stabilizing the system. Convergence analysis is provided.
Article
Engineering, Electrical & Electronic
Jikeng Lin, Kaiming Yuan, Lingfeng Wang
Summary: This study introduces a new adaptive sparse pseudospectral approximation method called NA-SPAM, which addresses the issues of inaccurate global error estimate and inefficiency in multi-output systems that were present in the original A-SPAM. Through two improvements, NA-SPAM demonstrates higher estimation accuracy and improved efficiency in calculating multi-output problems.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS
(2021)
Article
Water Resources
Tohid Erfani, Julien J. Harou
Summary: Dealing with uncertainty in water resource planning is complex due to the potential social and environmental costs of insufficient infrastructure. Multistage stochastic optimisation offers a solution to this challenge, but can be difficult and expensive for real systems. The 'Decision-rule' formulation approximates the multistage problem by introducing a series of rules that are functions of uncertainty and system state, ultimately impacting adaptive water resources planning.
ADVANCES IN WATER RESOURCES
(2021)
Article
Management
Yonatan Gur, Ahmadreza Momeni
Summary: This paper investigates the performance achieved by leveraging auxiliary information in sequential experiments and proposes effective algorithms. The study shows that upper confidence bound and Thompson sampling algorithms have good performance when the mapping between auxiliary observations and rewards is known, and auxiliary information can improve performance. When the mapping is unknown, an adaptive strategy is proposed to ensure near optimality, and better performance can be achieved by utilizing auxiliary observations in practical applications.
M&SOM-MANUFACTURING & SERVICE OPERATIONS MANAGEMENT
(2022)
Article
Management
Said Rahal, Dimitri J. Papageorgiou, Zukui Li
Summary: Decision rules provide a rich framework for solving multistage adaptive optimization problems, with recent literature showing the potential of using both linear and nonlinear decision rules. The study explores hybrid decision rules combining the benefits of the two classes, highlighting the trade-off between solution quality and computational cost. Unexpectedly, a linear decision rule was found to be superior to a more complex piecewise-linear decision rule in a simulator, emphasizing the importance of assessing decision rule quality within a simulator.
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH
(2021)
Article
Computer Science, Interdisciplinary Applications
Shaojie Tang
Summary: The goal of a sequential decision-making problem is to design an interactive policy that adaptively selects a group of items, each selection is based on the feedback from the past, to maximize the expected utility of selected items. This study proposes to study two variants of adaptive submodular optimization problems and introduces a new class of stochastic functions called worst-case submodular functions. Several applications of the theoretical results are also described.
INFORMS JOURNAL ON COMPUTING
(2022)