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
  • [41] Automated Benchmark Model Generators for Model-Based Diagnostic Inference
    Provan, Gregory
    Wang, Jun
    20TH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2007, : 513 - 518
  • [42] A Comparison of Two Sampling Inference Systems between Design-Based Inference and Model-Based Inference
    He Benlan
    Jin Yongjin
    Gong Hongyu
    RECENT ADVANCE IN STATISTICS APPLICATION AND RELATED AREAS, PTS 1 AND 2, 2011, : 1373 - +
  • [43] DESIGN-BASED AND MODEL-BASED INFERENCE IN SURVEY SAMPLING
    SARNDAL, CE
    SCANDINAVIAN JOURNAL OF STATISTICS, 1978, 5 (01) : 27 - 52
  • [44] Model-based diagnostics and probabilistic assumption-based reasoning
    Kohlas, J
    Anrig, B
    Haenni, R
    Monney, PA
    ARTIFICIAL INTELLIGENCE, 1998, 104 (1-2) : 71 - 106
  • [45] Model-based Clustering with Noise: Bayesian Inference and Estimation
    H. Bensmail
    J. J. Meulman
    Journal of Classification, 2003, 20 : 049 - 076
  • [46] Spline density estimation and inference with model-based penalties
    Shi, Jian
    Liu, Anna
    Wang, Yuedong
    JOURNAL OF NONPARAMETRIC STATISTICS, 2019, 31 (03) : 596 - 611
  • [47] Model-based clustering with noise: Bayesian inference and estimation
    Bensmail, H
    Meulman, JJ
    JOURNAL OF CLASSIFICATION, 2003, 20 (01) : 49 - 76
  • [48] Bayesian model-based inference of transcription factor activity
    Simon Rogers
    Raya Khanin
    Mark Girolami
    BMC Bioinformatics, 8
  • [49] A frequentist interpretation of probability for model-based inductive inference
    Aris Spanos
    Synthese, 2013, 190 : 1555 - 1585
  • [50] Sensitivity analysis practices: Strategies for model-based inference
    Saltelli, Andrea
    Ratto, Marco
    Tarantola, Stefano
    Campolongo, Francesca
    RELIABILITY ENGINEERING & SYSTEM SAFETY, 2006, 91 (10-11) : 1109 - 1125