Complexity of Derivative-Free Policy Optimization for Structured $\mathcal{H}_\infty$ Control

Authors: Xingang Guo, Darioush Keivan, Geir Dullerud, Peter Seiler, Bin Hu

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we present the numerical study to show the effectiveness of Algorithm 1. The left plot of Figure 1 displays the relative error trajectories of Algorithm 1 with an exact oracle for system dimensions nx = {10, 50, 100}.
Researcher Affiliation Academia Xingang Guo ECE, CSL UIUC xingang2@illinois.edu Darioush Keivan MSE, CSL UIUC dk12@illinois.edu Geir Dullerud MSE, CSL UIUC dullerud@illinois.edu Peter Seiler EECS UMich pseiler@umich.edu Bin Hu ECE, CSL UIUC binhu7@illinois.edu
Pseudocode Yes Algorithm 1: Derivative-Free Methods for Policy Optimization Problem (4)
Open Source Code No The paper does not provide an explicit statement or link for open-source code for the described methodology.
Open Datasets Yes In addition, we conducted a comparison of our derivative-free method with the model-based methods HIFOO and Hinfstruct, using several benchmark examples from COMPleib [41].
Dataset Splits No The paper does not specify training, validation, or test dataset splits. The research focuses on control synthesis rather than traditional machine learning dataset evaluation.
Hardware Specification Yes All the experiments are performed on a desktop computer with a 3.7 GHz Intel i5-9600K processor.
Software Dependencies No The paper mentions using MATLAB and specific packages (HIFOO, Hinfstruct) for comparison, but does not provide specific version numbers for the software dependencies used to run their experiments.
Experiment Setup Yes For Algorthm 1, we set η = 1 10 3, δ = 1 10 4, and ϵ = 1 10 3.