Deep Probabilistic Accelerated Evaluation: A Robust Certifiable Rare-Event Simulation Methodology for Black-Box Safety-Critical Systems

Arief, Mansur; Huang, Zhiyuan; Kumar, Guru Koushik Senthil; Bai, Yuanlu; He, Shengyi; Ding, Wenhao; Lam, Henry H N; Zhao, Ding

doi:10.48550/arxiv.2006.15722

Cited by 6 publications

(6 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[122] uses GMM for IS distribution and further analyzes the efficiency of GMM based IS distribution for random forest and neural network classifiers [123]. ReLUactivated deep neural network is considered in [124] to estimate the dangerous set and compute an IS estimator for a risk upper bound for Gaussian case, with a more general case presented in [54]. The Adaptive IS approach is used to construct adversarial environments to accelerate policy evaluation [125].…”

Section: Adversarial Policymentioning

confidence: 99%

A Survey on Safety-Critical Driving Scenario Generation -- A Methodological Perspective

Ding¹,

Xu²,

Arief³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Autonomous driving systems have witnessed a significant development during the past years thanks to the advance in machine learning-enabled sensing and decision-making algorithms. One critical challenge for their massive deployment in the real world is their safety evaluation. Most existing driving systems are still trained and evaluated on naturalistic scenarios collected from daily life or heuristically-generated adversarial ones. However, the large population of cars, in general, leads to an extremely low collision rate, indicating that the safety-critical scenarios are rare in the collected real-world data. Thus, methods to artificially generate scenarios become crucial to measure the risk and reduce the cost. In this survey, we focus on the algorithms of safety-critical scenario generation in autonomous driving. We first provide a comprehensive taxonomy of existing algorithms by dividing them into three categories: data-driven generation, adversarial generation, and knowledge-based generation. Then, we discuss useful tools for scenario generation, including simulation platforms and packages. Finally, we extend our discussion to five main challenges of current works -fidelity, efficiency, diversity, transferability, controllability -and research opportunities lighted up by these challenges.

show abstract

Section: Adversarial Policymentioning

confidence: 99%

A Survey on Safety-Critical Driving Scenario Generation -- A Methodological Perspective

Ding¹,

Xu²,

Arief³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…We benchmark the proposed Deep IS framework with two other accelerated evaluation methods. First is the Deep Probabilistic Accelerated Evaluation approach first proposed in [29], which we called Robust Deep IS here. The 'robustness' of this approach stems from its nature of estimating an upper-bound of the safety risk, instead of the true target, providing some level of hedging against unmeasurable risk.…”

Section: Benchmark #1: Robust Deep Ismentioning

confidence: 99%

“…The use of outer-approximation guarantees the theoretical efficiency of Robust Deep IS estimator when the dangerous set S γ is orthogonally monotone. The robustness of this estimator is due to its interesting use-case to estimate an upper bound μRobustDeepIS that is guaranteed to avoid underestimating the risk [29], with a property that E μRobustDeepIS ≥ µ . The tuning method for κ provides a way to minimize the conservativeness of the estimator, hence to some extent, can be viewed as a minimax estimator for the risk (i.e.…”

Section: Benchmark #1: Robust Deep Ismentioning

confidence: 99%

“…We add an additional benchmark, the Iterative Robust Deep IS (Alg. 3), as an improvement to the risk upper-bound estimator proposed in [29]. In addition, we conduct more experiments: AV driving controller evaluation under cut-in-tailgate scenario and an ablation study of the traffic sign classifier evaluation, varying the rarity parameter to better study the efficiency trend of Deep IS method against rarer failure cases.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Test Against High-Dimensional Uncertainties: Accelerated Evaluation of Autonomous Vehicles with Deep Importance Sampling

Arief¹,

Cen²,

Liu³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Evaluating the performance of autonomous vehicles (AV) and their complex subsystems to high precision under naturalistic circumstances remains a challenge, especially when the failure or dangerous cases are rare. Rarity does not only require an enormous sample size for a naive method to achieve high confidence estimation, but it also causes dangerous underestimation of the true failure rate and it is extremely hard to detect. Meanwhile, the state-of-the-art approach that comes with a correctness guarantee can only compute an upper bound for the failure rate under certain conditions, which could limit its practical uses. In this work, we present Deep Importance Sampling (Deep IS) framework that utilizes a deep neural network to obtain an efficient IS that is on-par with the state-of-the-art, capable of reducing the required sample size 43 times smaller than the naive sampling method to achieve 10% relative error and while producing an estimate that is much less conservative. Our high-dimensional experiment estimating the misclassification rate of one of the state-of-the-art traffic sign classifiers further reveals that this efficiency still holds true even when the target is very small, achieving over 600 times efficiency boost. This highlights the potential of Deep IS in providing a precise estimate even against highdimensional uncertainties.

show abstract

“…Motivated by the rare nature of safety-critical events in lab and field tests [76,57], there is a rich literature on rare event probability estimation [7,14,66,54] devoting to maximize the sampling efficiency, among which importance sampling [56,48,55,4] and multi-level splitting [29,67,8] stand as two powerhouses. However, few literature addresses the issue under the setting of Markov decision process (MDP).…”

Section: Related Workmentioning

confidence: 99%

Accelerated Policy Evaluation: Learning Adversarial Environments with Adaptive Importance Sampling

Xu¹,

Huang²,

Li³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

The evaluation of rare but high-stakes events remains one of the main difficulties in obtaining reliable policies from intelligent agents, especially in large or continuous state/action spaces where limited scalability enforces the use of a prohibitively large number of testing iterations. On the other hand, a biased or inaccurate policy evaluation in a safety-critical system could potentially cause unexpected catastrophic failures during deployment. In this paper, we propose the Accelerated Policy Evaluation (APE) method, which simultaneously uncovers rare events and estimates the rare event probability in Markov decision processes. The APE method treats the environment nature as an adversarial agent and learns towards, through adaptive importance sampling, the zero-variance sampling distribution for the policy evaluation. Moreover, APE is scalable to large discrete or continuous spaces by incorporating function approximators. We investigate the convergence properties of proposed algorithms under suitable regularity conditions. Our empirical studies show that APE estimates rare event probability with a smaller variance while only using orders of magnitude fewer samples compared to baseline methods in both multi-agent and single-agent environments.Preprint. Under review.

show abstract

Deep Probabilistic Accelerated Evaluation: A Robust Certifiable Rare-Event Simulation Methodology for Black-Box Safety-Critical Systems

Cited by 6 publications

References 33 publications

A Survey on Safety-Critical Driving Scenario Generation -- A Methodological Perspective

A Survey on Safety-Critical Driving Scenario Generation -- A Methodological Perspective

Test Against High-Dimensional Uncertainties: Accelerated Evaluation of Autonomous Vehicles with Deep Importance Sampling

Accelerated Policy Evaluation: Learning Adversarial Environments with Adaptive Importance Sampling

Contact Info

Product

Resources

About