![[2404.12097] MPC of Uncertain Nonlinear Systems with Meta-Learning for Fast Adaptation of Neural Predictive Models](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2404.12097] MPC of Uncertain Nonlinear Systems with Meta-Learning for Fast Adaptation of Neural Predictive Models
Summary
This paper presents a novel approach to model predictive control (MPC) for uncertain nonlinear systems using a neural state-space model and meta-learning for rapid adaptation.
Why It Matters
The study addresses the challenge of controlling uncertain nonlinear systems, which are prevalent in various engineering applications. By utilizing meta-learning, the proposed method enhances the efficiency of model adaptation, potentially leading to improved performance in real-time control scenarios. This has significant implications for fields such as robotics and automation, where quick and accurate decision-making is crucial.
Key Takeaways
- Introduces a neural state-space model (NSSM) for nonlinear system approximation.
- Utilizes an implicit model-agnostic meta-learning (iMAML) framework for fast adaptation.
- Demonstrates improved control performance in numerical examples compared to baseline methods.
- Focuses on reducing storage complexity and approximations in meta-learning.
- Highlights the importance of leveraging data from similar systems for effective training.
Electrical Engineering and Systems Science > Systems and Control arXiv:2404.12097 (eess) [Submitted on 18 Apr 2024 (v1), last revised 25 Feb 2026 (this version, v2)] Title:MPC of Uncertain Nonlinear Systems with Meta-Learning for Fast Adaptation of Neural Predictive Models Authors:Jiaqi Yan, Ankush Chakrabarty, Alisa Rupenyan, John Lygeros View a PDF of the paper titled MPC of Uncertain Nonlinear Systems with Meta-Learning for Fast Adaptation of Neural Predictive Models, by Jiaqi Yan and 3 other authors View PDF HTML (experimental) Abstract:In this paper, we consider the problem of reference tracking in uncertain nonlinear systems. A neural State-Space Model (NSSM) is used to approximate the nonlinear system, where a deep encoder network learns the nonlinearity from data, and a state-space component captures the temporal relationship. This transforms the nonlinear system into a linear system in a latent space, enabling the application of model predictive control (MPC) to determine effective control actions. Our objective is to design the optimal controller using limited data from the \textit{target system} (the system of interest). To this end, we employ an implicit model-agnostic meta-learning (iMAML) framework that leverages information from \textit{source systems} (systems that share similarities with the target system) to expedite training in the target system and enhance its control performance. The framework consists of two phases: the (offine) meta-training phase le...
