Model-based Validation as Probabilistic Inference

被引:0
|
作者
Delecki, Harrison [1 ]
Corso, Anthony [1 ]
Kochenderfer, Mykel J. [1 ]
机构
[1] Stanford Univ, Dept Aeronaut & Astronaut, Stanford, CA 94305 USA
基金
美国国家科学基金会;
关键词
safety validation; Bayesian inference;
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
Estimating the distribution over failures is a key step in validating autonomous systems. Existing approaches focus on finding failures for a small range of initial conditions or make restrictive assumptions about the properties of the system under test. We frame estimating the distribution over failure trajectories for sequential systems as Bayesian inference. Our model-based approach represents the distribution over failure trajectories using rollouts of system dynamics and computes trajectory gradients using automatic differentiation. Our approach is demonstrated in an inverted pendulum control system, an autonomous vehicle driving scenario, and a partially observable lunar lander. Sampling is performed using an off-the-shelf implementation of Hamiltonian Monte Carlo with multiple chains to capture multimodality and gradient smoothing for safe trajectories. In all experiments, we observed improvements in sample efficiency and parameter space coverage compared to black-box baseline approaches. This work is open sourced.
引用
收藏
页数:13
相关论文
共 50 条
  • [1] Model-based kernel sum rule: kernel Bayesian inference with probabilistic models
    Nishiyama, Yu
    Kanagawa, Motonobu
    Gretton, Arthur
    Fukumizu, Kenji
    [J]. MACHINE LEARNING, 2020, 109 (05) : 939 - 972
  • [2] Model-based kernel sum rule: kernel Bayesian inference with probabilistic models
    Yu Nishiyama
    Motonobu Kanagawa
    Arthur Gretton
    Kenji Fukumizu
    [J]. Machine Learning, 2020, 109 : 939 - 972
  • [3] MODEL-BASED INFERENCE IN CHARME
    PESCH, E
    DREXL, A
    KOLEN, A
    [J]. OR SPEKTRUM, 1994, 16 (03) : 193 - 202
  • [4] Probabilistic model-based diagnosis
    Ibargüengoytia, PH
    Sucar, LE
    Morales, E
    [J]. MICAI 2000: ADVANCES IN ARTIFICIAL INTELLIGENCE, PROCEEDINGS, 2000, 1793 : 687 - 698
  • [5] Inference in model-based cluster analysis
    Halima Bensmail
    Gilles Celeux
    Adrian E. Raftery
    Christian P. Robert
    [J]. Statistics and Computing, 1997, 7 : 1 - 10
  • [6] Model-Based Inference of Synaptic Transmission
    Bykowska, Ola
    Gontier, Camille
    Sax, Anne-Lene
    Jia, David W.
    Montero, Milton Llera
    Bird, Alex D.
    Houghton, Conor
    Pfister, Jean-Pascal
    Costa, Rui Ponte
    [J]. FRONTIERS IN SYNAPTIC NEUROSCIENCE, 2019, 11
  • [7] Inference in model-based cluster analysis
    Bensmail, H
    Celeux, G
    Raftery, AE
    Robert, CP
    [J]. STATISTICS AND COMPUTING, 1997, 7 (01) : 1 - 10
  • [8] A methodology for probabilistic model-based prognosis
    Lorton, A.
    Fouladirad, M.
    Grall, A.
    [J]. EUROPEAN JOURNAL OF OPERATIONAL RESEARCH, 2013, 225 (03) : 443 - 454
  • [9] Probabilistic Model-based Background Subtraction
    Krüger, V
    Anderson, J
    Prehn, T
    [J]. IMAGE ANALYSIS AND PROCESSING - ICIAP 2005, PROCEEDINGS, 2005, 3617 : 180 - 187
  • [10] Probabilistic model-based background subtraction
    Krüger, V
    Anderson, J
    Prehn, T
    [J]. IMAGE ANALYSIS, PROCEEDINGS, 2005, 3540 : 567 - 576