Design of adaptive controllers by means of PPO algorithm using MATLAB

被引：0

作者：

Radojicic, Veljko ^{[1
]}

Rapaic, Milan R. ^{[1
]}

机构：

[1] Univ Novi Sad, Dept Comp & Control Engn, Fac Tech Sci, Novi Sad, Serbia

来源：

2025 24TH INTERNATIONAL SYMPOSIUM INFOTEH-JAHORINA, INFOTEH | 2025年

关键词：

PPO; policy; agent; controller;

D O I：

10.1109/INFOTEH64129.2025.10959267

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

This paper investigates reinforcement learning (RL) as a practical framework for achieving optimal adaptive control across several simple dynamical system models. All experiments were conducted using the Proximal Policy Optimization (PPO) algorithm, implemented within the MATLAB Reinforcement Learning Toolbox. The primary focus of this study is to explore how the learning process can be empirically designed to sufficiently excite the system dynamics and obtain a sufficiently robust controller.

引用

页数：6

共 15 条

[1]

Costa P., 2022, ICLR Blog

[2]

Gebotys B, 2022, Arxiv, DOI [arXiv:2201.09104, 10.48550/arXiv.2201.09104, DOI 10.48550/ARXIV.2201.09104]

[3]

Lu Shuai, arXiv, DOI DOI 10.48550/ARXIV.2011.05525

[4]

Mnih V, 2013, Arxiv, DOI arXiv:1312.5602

[5]

Petrazzini Irving G. B., arXiv, DOI [10.48550/arXiv.2111.02202, DOI 10.48550/ARXIV.2111.02202]

[6]

Schulman J., 2015, arXiv, DOI [arXiv:1502.05477, DOI 10.48550/ARXIV.1502.05477]

[7]

Schulman J, 2017, Arxiv, DOI arXiv:1707.06347

[8]

Suilen M, 2024, Arxiv, DOI [arXiv:2411.11451, 10.48550/arXiv.2411.11451, DOI 10.48550/ARXIV.2411.11451]

[9]

Sutton RS, 2018, ADAPT COMPUT MACH LE, P1

[10]

The MathWorks Inc., Reinforcement Learning Designer

← 1 2 →