Model-Agnostic Meta-Policy Optimization via Zeroth-Order Estimation: A Linear Quadratic Regulator Perspective
Yunian Pan, Tao Li, Quanyan Zhu · 2025
Read Model-Agnostic Meta-Policy Optimization via Zeroth-Order Estimation: A Linear Quadratic Regulator Perspective by Yunian Pan, Tao Li, Quanyan Zhu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-00385.txt.
Description
Yunian Pan, Tao Li, and Quanyan Zhu study how a controller can learn to adapt across a family of similar machines. Meta-learning trains a starting policy that adjusts fast to each new task. Their setting is the linear quadratic regulator, a clean model of control with steady rules. Real systems differ in detail, so one fixed policy rarely fits all.
Their method estimates gradients from raw trials alone, with no need for costly second-order terms. The meta-goal keeps the shape of the original cost when the tasks are truly learnable as a family. That shape lets plain descent work without extra projections. The authors prove stability and convergence under stated sample sizes. A worked example backs the theory. The result is a principled route to adaptive control for mixed fleets of systems.
Catalog record
| CREATOR | Yunian Pan, Tao Li, Quanyan Zhu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 92 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-00385