Efficient Autonomy Validation in Simulation with Adaptive Stress Testing

Koren, Mark; Kochenderfer, Mykel J.

doi:10.1109/itsc.2019.8917403

Cited by 32 publications

(24 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[62] generates a scenario controlling a pedestrian to cross the road. [63] improves the last paper by using LSTM to generate initial conditions and actions in each step. Instead of defining heuristic reward functions, [64] leverage the Go-Explore framework to find failure cases.…”

Section: Adversarial Policymentioning

confidence: 99%

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

Ding¹,

Xu²,

Arief³

et al. 2022

Preprint

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

View full text Add to dashboard Cite

show abstract

“…Previous papers have presented solvers based on Monte Carlo tree search [10], deep reinforcement learning (DRL) [11], and go-explore [12]. In this paper, we will use DRL and the BA, with background provided for those unfamiliar with either approach.…”

Section: B Formulationmentioning

confidence: 99%

Finding Failures in High-Fidelity Simulation using Adaptive Stress Testing and the Backward Algorithm

Koren¹,

Nassar²,

Kochenderfer³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Validating the safety of autonomous systems generally requires the use of high-fidelity simulators that adequately capture the variability of real-world scenarios. However, it is generally not feasible to exhaustively search the space of simulation scenarios for failures. Adaptive stress testing (AST) is a method that uses reinforcement learning to find the most likely failure of a system. AST with a deep reinforcement learning solver has been shown to be effective in finding failures across a range of different systems. This approach generally involves running many simulations, which can be very expensive when using a high-fidelity simulator. To improve efficiency, we present a method that first finds failures in a low-fidelity simulator. It then uses the backward algorithm, which trains a deep neural network policy using a single expert demonstration, to adapt the low-fidelity failures to high-fidelity. We have created a series of autonomous vehicle validation case studies that represent some of the ways low-fidelity and highfidelity simulators can differ, such as time discretization. We demonstrate in a variety of case studies that this new AST approach is able to find failures with significantly fewer highfidelity simulation steps than are needed when just running AST directly in high-fidelity. As a proof of concept, we also demonstrate AST on NVIDIA's DriveSim simulator, an industry state-of-the-art high-fidelity simulator for finding failures in autonomous vehicles.

show abstract

“…where Dist(s) is some measure of the simulator's closeness to a failure, and α and β scale the penalty term given when a terminal state is reached that is not a failure [21]. In practice α and β are very large to encourage the simulator to find a failure before optimizing the action sequence to reach failure.…”

Section: B Reward Functionmentioning

confidence: 99%

“…In previous work with AST, the action space has been lowdimensional [20], [21]. However, when the action space is a high-dimensional image, previous AST algorithms will not perform well.…”

Section: Action Spacementioning

confidence: 99%

“…This work focuses on validation of image-based neural network controllers. Existing work on validation of complex systems has led to the development of adaptive stress testing (AST), which uses reinforcement learning to find the most likely ways systems fail [20], [21]. However, existing work with AST has only considered low-dimensional problems.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Validation of Image-Based Neural Network Controllers through Adaptive Stress Testing

Julian

Lee

Kochenderfer

2020

2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC)

Self Cite

View full text Add to dashboard Cite

Neural networks have become state-of-the-art for computer vision problems because of their ability to efficiently model complex functions from large amounts of data. While neural networks can be shown to perform well empirically for a variety of tasks, their performance is difficult to guarantee. Neural network verification tools have been developed that can certify robustness with respect to a given input image; however, for neural network systems used in closed-loop controllers, robustness with respect to individual images does not address multi-step properties of the neural network controller and its environment. Furthermore, neural network systems interacting in the physical world and using natural images are operating in a black-box environment, making formal verification intractable. This work combines the adaptive stress testing (AST) framework with neural network verification tools to search for the most likely sequence of image disturbances that cause the neural network controlled system to reach a failure. An autonomous aircraft taxi application is presented, and results show that the AST method finds failures with more likely image disturbances than baseline methods. Further analysis of AST results revealed an explainable cause of the failure, giving insight into the problematic scenarios that should be addressed.

show abstract

Efficient Autonomy Validation in Simulation with Adaptive Stress Testing

Cited by 32 publications

References 14 publications

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

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

Finding Failures in High-Fidelity Simulation using Adaptive Stress Testing and the Backward Algorithm

Validation of Image-Based Neural Network Controllers through Adaptive Stress Testing

Contact Info

Product

Resources

About