Learning Min-norm Stabilizing Control Laws for Systems with Unknown Dynamics

被引:0
|
作者
Westenbroek, Tyler [1 ]
Castaneda, Fernando [2 ]
Agrawal, Ayush [2 ]
Sastry, S. Shankar [1 ]
Sreenath, Koushil [2 ]
机构
[1] Univ Calif Berkeley, Dept Elect Engn & Comp Sci, Berkeley, CA 94720 USA
[2] Univ Calif Berkeley, Dept Mech Engn, Berkeley, CA 94720 USA
来源
2020 59TH IEEE CONFERENCE ON DECISION AND CONTROL (CDC) | 2020年
基金
美国国家科学基金会;
关键词
TO-STATE STABILITY; ADAPTIVE-CONTROL; LYAPUNOV;
D O I
10.1109/cdc42340.2020.9304118
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 introduces a framework for learning a minimum-norm stabilizing controller for a system with unknown dynamics using model-free policy optimization methods. The approach begins by first designing a Control Lyapunov Function (CLF) for a (possibly inaccurate) dynamics model for the system, along with a function which specifies a minimum acceptable rate of energy dissipation for the CLF at different points in the state-space. Treating the energy dissipation condition as a constraint on the desired closed-loop behavior of the real-world system, we use penalty methods to formulate an unconstrained optimization problem over the parameters of a learned controller, which can be solved using model-free policy optimization algorithms using data collected from the plant. We discuss when the optimization learns a stabilizing controller for the real world system and derive conditions on the structure of the learned controller which ensure that the optimization is strongly convex, meaning the globally optimal solution can be found reliably. We validate the approach in simulation, first for a double pendulum, and then generalize the framework to learn stable walking controllers for underactuated bipedal robots using the Hybrid Zero Dynamics framework. By encoding a large amount of structure into the learning problem, we are able to learn stabilizing controllers for both systems with only minutes or even seconds of training data.
引用
收藏
页码:737 / 744
页数:8
相关论文
共 50 条
  • [11] Pointwise Minimum Norm Control Laws for Hybrid Systems
    Sanfelice, Ricardo G.
    2013 IEEE 52ND ANNUAL CONFERENCE ON DECISION AND CONTROL (CDC), 2013, : 2665 - 2670
  • [12] PARAMETRIZATION OF NONLINEAR STABILIZING CONTROL LAWS FOR BIFURCATIONS AND CRITICAL SYSTEMS
    FU, JH
    ABED, EH
    IEEE INTERNATIONAL CONFERENCE ON SYSTEMS ENGINEERING ///, 1989, : 437 - 441
  • [13] Design of continuous stabilizing control laws for nonlinear triangular systems
    Aranda-Bricaire, E
    Celikovsky, S
    SYSTEM STRUCTURE AND CONTROL 1998 (SSC'98), VOLS 1 AND 2, 1998, : 77 - 82
  • [14] DESIGN OF STABILIZING CONTROL LAWS FOR SMOOTH NONLINEAR-SYSTEMS
    MICHALSKA, H
    MAYNE, DQ
    INTERNATIONAL JOURNAL OF CONTROL, 1991, 53 (03) : 541 - 558
  • [15] Myopic Control of Systems with Unknown Dynamics
    Ornik, Melkior
    Carr, Steven
    Israel, Arie
    Topcu, Ufuk
    2019 AMERICAN CONTROL CONFERENCE (ACC), 2019, : 1064 - 1071
  • [16] Learning Models for Shared Control of Human-Machine Systems with Unknown Dynamics
    Broad, Alexander
    Murphey, Todd
    Argall, Brenna
    ROBOTICS: SCIENCE AND SYSTEMS XIII, 2017,
  • [17] Continuous Action Reinforcement Learning for Control-Affine Systems with Unknown Dynamics
    Aleksandra Faust
    Peter Ruymgaart
    Molly Salman
    Rafael Fierro
    Lydia Tapia
    IEEE/CAAJournalofAutomaticaSinica, 2014, 1 (03) : 323 - 336
  • [18] Continuous action reinforcement learning for control-affine systems with unknown dynamics
    Faust, Aleksandra
    Ruymgaart, Peter
    Salman, Molly
    Fierro, Rafael
    Tapia, Lydia
    IEEE/CAA Journal of Automatica Sinica, 2014, 1 (03) : 323 - 336
  • [19] A stabilizing reinforcement learning approach for sampled systems with partially unknown models
    Beckenbach, Lukas
    Osinenko, Pavel
    Streif, Stefan
    INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL, 2024, 34 (18) : 12389 - 12412
  • [20] Iterative Learning Control Laws with Full Dynamics
    Hladowski, Lukasz
    Chen, Yiyang
    Nowicka, Weronika
    Galkowski, Krzysztof
    Rogers, Eric
    2016 AMERICAN CONTROL CONFERENCE (ACC), 2016, : 6369 - 6374