Proceedings Paper
Automation & Control Systems
Wei Zheng, Hai Lin
Summary: The paper extends the traditional POMDP to VAR-POMDP, discusses solving the issue of temporal correlation among continuous observations, and proposes a double point-based value iteration algorithm to address this problem.
2021 AMERICAN CONTROL CONFERENCE (ACC)
(2021)
Article
Engineering, Mechanical
Zhe Chen, Wenqian Xue, Ning Li, Bosen Lian, Frank L. Lewis
Summary: This paper proposes a completely model-free reinforcement learning (RL) method for solving the finite-horizon two-player zero-sum game problem of continuous-time nonlinear systems. By defining a novel Z-function, introducing a model-based RL policy iteration framework, and applying integral RL and iterative learning control techniques, the solution seeking and system dynamics requirement are further simplified, leading to improved algorithm efficiency and reduced model dependency.
NONLINEAR DYNAMICS
(2022)
Article
Operations Research & Management Science
Dongmei Yu, Cairong Chen, Deren Han
Summary: This article presents a modified fixed point iteration method for solving absolute value equations, which improves efficiency and shows linear convergence under certain conditions. Numerical results demonstrate its superiority compared to the original method in specific cases.
Article
Automation & Control Systems
Wei Zheng, Hai Lin
Summary: This paper introduces a VAR-POMDP model which extends the traditional POMDP model and proposes a feasible planning algorithm. The VAR-POMDP model can be solved by approximating the exact value function using a class of piece-wise linear functions within the dynamic programming framework.
IEEE CONTROL SYSTEMS LETTERS
(2022)
Article
Automation & Control Systems
Hassan Hmedi, Ari Arapostathis, Guodong Pang
Summary: In this paper, a multiplicative relative value iteration algorithm (RVI) for infinite-horizon risk-sensitive control of diffusions in Rd is studied. The author proves that the RVI algorithm converges to the solution of the multiplicative HJB equation within a neighborhood of the solution (local convergence) when the diffusion is positive recurrent. Under the assumption of blanket exponential ergodicity, it is also shown that the RVI algorithm converges globally to the solution of the multiplicative HJB equation from any positive initial condition. This paper revisits the problem without assuming blanket conditions, instead assuming a near-monotone running cost and a structural assumption relating the running cost function to the solution of the multiplicative HJB equation. It is shown that this structural assumption implies the existence of a control under which the ground state diffusion is exponentially ergodic, and a global convergence result of the multiplicative VI/RVI algorithms is established, extending the results in Arapostathis and Borkar (2020).
SYSTEMS & CONTROL LETTERS
(2023)
Article
Mathematics, Applied
S. Thenmozhi, M. Marudai
Summary: In this paper, a novel approach was introduced to solve the non-linear fourth order boundary value problem using integral operator equation. Through three illustrations, it was demonstrated that S-iteration for contraction operator shows faster convergence than Krasnoselskii-Mann's iteration based on residual or absolute error calculations.
JOURNAL OF APPLIED MATHEMATICS AND COMPUTING
(2022)
Article
Computer Science, Artificial Intelligence
Shubham Pateria, Budhitama Subagdja, Ah-Hwee Tan, Chai Quek
Summary: This article proposes a novel subgoal graph-based planning method called LSGVP, which addresses the challenge of learning to reach long-horizon goals in spatial traversal tasks for autonomous agents. LSGVP uses a subgoal discovery heuristic based on cumulative reward and automatically prunes the learned subgoal graph to remove erroneous connections. It achieves higher cumulative positive rewards and goal-reaching success rates compared to other subgoal sampling or discovery heuristics.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Ziyu Lin, Jingliang Duan, Shengbo Eben Li, Haitong Ma, Jie Li, Jianyu Chen, Bo Cheng, Jun Ma
Summary: The research addresses the challenge of solving the finite-horizon HJB equation, proposes a new algorithm, and validates its effectiveness through simulations.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Mathematics, Applied
Xu Li, Yi-Xin Li, Yan Dou
Summary: In this paper, we propose a shift-splitting fixed point iteration method (FPI-SS) for solving large sparse generalized absolute value equations (GAVEs). Various convergence conditions of the FPI-SS method are presented. Numerical experiments demonstrate that the FPI-SS method outperforms other methods in terms of computing efficiency.
NUMERICAL ALGORITHMS
(2023)
Article
Automation & Control Systems
Zhe Chen, Wenqian Xue, Ning Li, Frank L. Lewis
Summary: This article introduces three novel time-varying policy iteration algorithms for finite-horizon optimal control problem of continuous-time affine nonlinear systems, including model-based and partially model-free methods, and provides analysis on the convergence, stability, and optimality of each algorithm.
INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL
(2022)
Article
Mathematics
Mujahid Abbas, Rizwan Anjum, Vasile Berinde
Summary: This paper aims to define two new classes of mappings, demonstrate the existence and iterative approximation of their fixed points, and show the equivalence of Ishikawa, Mann, and Krasnoselskij iteration methods for such mappings. Additionally, applications of these results to solve split feasibility and variational inequality problems are provided.
Article
Mathematics, Applied
Heng Zhang
Summary: This paper proposes a novel adaptive dynamic programming (ADP)-based model-free policy iteration (PI) algorithm to solve an infinite-horizon continuous-time linear quadratic stochastic (LQS) optimal control problem, which includes both control and state variables in the diffusion term of system dynamics. By using Ito's lemma and expectations, a relationship among the state trajectory, control input, and matrices to be solved is described. The ADP-based model-free algorithm is then developed to approximate the optimal control from collected data without requiring information about all system coefficient matrices. Convergence analysis is provided under mild conditions, and numerical examples demonstrate the effectiveness of the proposed algorithm.
JOURNAL OF APPLIED MATHEMATICS AND COMPUTING
(2023)
Article
Computer Science, Information Systems
Wenjun Xiong, Daniel W. C. Ho, Shifan Wen
Summary: This study analyzes the finite-iteration tracking of discrete networks by designing a new periodic ILC strategy using the FlexRay communication protocol. The approach reduces communication channel bandwidth load and improves performance. It provides a new method for iterative learning in network control design and shows better performance compared to traditional ILC schemes.
INFORMATION SCIENCES
(2021)
Article
Mathematics, Applied
Hongyu Wu, Shuhuang Xiang
Summary: A new constraint preconditioner is proposed for non-Hermitian generalized saddle point problems, constructed based on the PGSS iteration method. The invertibility condition and convergence properties of the new preconditioner are analyzed in detail, and its effectiveness is illustrated through numerical experiments.
APPLIED MATHEMATICS AND COMPUTATION
(2021)
Article
Mathematics, Applied
Stefania Ragni
Summary: An optimal control model governed by parabolic equations is analyzed through the formulation of the optimality system. The uniqueness of the solution is proven, and a constructive approximation method is provided. The convergence of iterative schemes and the use of exponential integrators in PDE-constrained optimization are investigated. Numerical results demonstrate the effectiveness of the proposed approach.
JOURNAL OF MATHEMATICAL ANALYSIS AND APPLICATIONS
(2022)