An Accelerated Augmented Gradient Neural Network for Constrained Time-Varying Nonlinear Optimization
- 1 November 2026
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 37 (11) , 6129-6139
Abstract
The constrained time-varying nonlinear optimization (CTVNO) problems with the linear equality and inequality constraints have attracted increasing attention in recent years. Although some models have been developed to address these problems, they still suffer from two critical limitations: complex model architecture and slow transient response. These issues restrict their practical application in engineering. To overcome these drawbacks, an accelerated augmented gradient neural network (AAGNN) is proposed. By integrating the error function with energy function gradients, the AAGNN efficiently exploits the essential information of the original problem, thereby reducing the complexity and accelerating convergence. The theoretical analysis demonstrates that the proposed AAGNN excels in achieving high-precision solutions. This stems from its ability to rapidly eliminate large errors through the exponential convergence. Comparative numerical experiments across four benchmarks demonstrate that the AAGNN consistently achieves a lower residual error and faster convergence than existing models. Furthermore, application tests through robotic manipulator trajectory tracking control and portfolio selection of marketed securities confirm the model’s superior practical applicability in real-world scenarios.Keywords
Funding Information
- National Key Research and Development Program of China (2021ZD0201300)
- Major program of National natural science foundation of China (62595802)
- Foundation for Outstanding Research Groups of Hubei Province of China (2025AFA012)
- 111 Project on Computational Intelligence and Intelligent Control (B18024)
- National Natural Science Foundation of China (625B2074)
- Fundamental Research Funds for the Central Universities (2025JYCXJJ071)
This publication has 40 references indexed in Scilit:
- An activated variable parameter gradient‐based neural network for time‐variant constrained quadratic programming and its applicationsCAAI Transactions on Intelligence Technology, 2023
- A multi-constrained zeroing neural network for time-dependent nonlinear optimization with application to mobile robot tracking controlNeurocomputing, 2021
- Time-Varying Convex Optimization: Time-Structured Algorithms and ApplicationsProceedings of the IEEE, 2020
- Iterative methods for the delay Lyapunov equation with T-Sylvester preconditioningApplied Numerical Mathematics, 2019
- Zeroing neural networks: A surveyNeurocomputing, 2017
- Distributed Continuous-Time Convex Optimization With Time-Varying Cost FunctionsIEEE Transactions on Automatic Control, 2016
- Continuous and discrete Zhang dynamics for real-time varying nonlinear optimizationNumerical Algorithms, 2015
- Opportunities and challenges for a sustainable energy futureNature, 2012
- A non-interior implicit smoothing approach to complementarity problems for frictionless contactsComputer Methods in Applied Mechanics and Engineering, 2011
- Numerical Optimization2006