A Generalizable Physics-informed Learning Framework for Risk Probability Estimation

被引:0
|
作者
Wang, Zhuoyuan [1 ]
Nakahira, Yorie [1 ]
机构
[1] Carnegie Mellon Univ, Elect & Comp Engn Dept, Pittsburgh, PA 15213 USA
关键词
Stochastic safe control; physics-informed learning; risk probability estimation; SAFE EXPLORATION; REACHABILITY;
D O I
暂无
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Accurate estimates of long-term risk probabilities and their gradients are critical for many stochastic safe control methods. However, computing such risk probabilities in real-time and in unseen or changing environments is challenging. Monte Carlo (MC) methods cannot accurately evaluate the probabilities and their gradients as an infinitesimal devisor can amplify the sampling noise. In this paper, we develop an efficient method to evaluate the probabilities of long-term risk and their gradients. The proposed method exploits the fact that long-term risk probability satisfies certain partial differential equations (PDEs), which characterize the neighboring relations between the probabilities, to integrate MC methods and physics-informed neural networks. We provide theoretical guarantees of the estimation error given certain choices of training configurations. Numerical results show the proposed method has better sample efficiency, generalizes well to unseen regions, and can adapt to systems with changing parameters. The proposed method can also accurately estimate the gradients of risk probabilities, which enables first- and second-order techniques on risk probabilities to be used for learning and control.
引用
收藏
页数:13
相关论文
共 50 条
  • [31] Physics-Informed Explainable Continual Learning on Graphs
    Peng, Ciyuan
    Tang, Tao
    Yin, Qiuyang
    Bai, Xiaomei
    Lim, Suryani
    Aggarwal, Charu C.
    IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2024, 35 (09) : 11761 - 11772
  • [32] PHYSICS-INFORMED MULTITASK LEARNING FOR MATERIAL DEVELOPMENT
    Wu, Yulun
    Li, Yumeng
    PROCEEDINGS OF ASME 2023 AEROSPACE STRUCTURES, STRUCTURAL DYNAMICS, AND MATERIALS CONFERENCE, SSDM2023, 2023,
  • [33] Physics-informed deep learning for digital materials
    Zhang, Zhizhou
    Gu, Grace X.
    THEORETICAL AND APPLIED MECHANICS LETTERS, 2021, 11 (01)
  • [34] A Taxonomic Survey of Physics-Informed Machine Learning
    Pateras, Joseph
    Rana, Pratip
    Ghosh, Preetam
    APPLIED SCIENCES-BASEL, 2023, 13 (12):
  • [35] Physics-informed deep learning for digital materials
    Zhizhou Zhang
    Grace X Gu
    Theoretical & Applied Mechanics Letters, 2021, 11 (01) : 52 - 57
  • [36] A novel physics-informed framework for reconstruction of structural defects
    Li, Qi
    Liu, Fushun
    Wang, Bin
    Liu, D. Z.
    Qian, Zhenghua
    APPLIED MATHEMATICS AND MECHANICS-ENGLISH EDITION, 2022, 43 (11) : 1717 - 1730
  • [37] A novel physics-informed framework for reconstruction of structural defects
    Qi Li
    Fushun Liu
    Bin Wang
    D. Z. Liu
    Zhenghua Qian
    Applied Mathematics and Mechanics, 2022, 43 : 1717 - 1730
  • [38] A PHYSICS-INFORMED ACTION SELECTION FRAMEWORK FOR ROBOTIC HEATING
    Dhanaraj, Neel
    Manyar, Omey M.
    Krishnan, Vihan
    Gupta, Satyandra K.
    PROCEEDINGS OF ASME 2023 INTERNATIONAL DESIGN ENGINEERING TECHNICAL CONFERENCES AND COMPUTERS AND INFORMATION IN ENGINEERING CONFERENCE, IDETC-CIE2023, VOL 2, 2023,
  • [39] A novel physics-informed framework for reconstruction of structural defects
    Qi LI
    Fushun LIU
    Bin WANG
    D.Z.LIU
    Zhenghua QIAN
    Applied Mathematics and Mechanics(English Edition), 2022, 43 (11) : 1717 - 1730
  • [40] A physics-informed deep reinforcement learning framework for autonomous steel frame structure design
    Fu, Bochao
    Gao, Yuqing
    Wang, Wei
    COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING, 2024, : 3125 - 3144