Dense reinforcement learning for safety validation of autonomous vehicles

Feng, Shuo; Sun, Haitian; Yan, Xintao; Zhu, Haojie; Zou, Zhengxia; Shen, Shengyin; Liu, Henry

doi:10.1038/s41586-023-05732-2

Cited by 171 publications

(32 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The first one is how to build a high-fidelity simulation environment and the second one is how to develop an accelerated testing methodology that can evaluate the AV performance accurately and efficiently. This work is focusing on the first problem, where the high-fidelity simulator is a prerequisite and foundation for simulation-based AV testing applications 2 , 4 , 52 . In our prior works 2 , 4 , we have developed accelerated testing methodologies that use a purely data-driven NDE model, which can be replaced with the high-fidelity NeuralNDE.…”

Section: Discussionmentioning

confidence: 99%

“…This work is focusing on the first problem, where the high-fidelity simulator is a prerequisite and foundation for simulation-based AV testing applications 2 , 4 , 52 . In our prior works 2 , 4 , we have developed accelerated testing methodologies that use a purely data-driven NDE model, which can be replaced with the high-fidelity NeuralNDE. We should note that human-driven vehicles might behave differently if they are interacting with an AV 53 .…”

Section: Discussionmentioning

confidence: 99%

“…Autonomous driving technologies are revolutionizing the future of transportation systems in unprecedented ways and speeds. However, safety remains the key challenge for the development and deployment of highly automated driving systems 1 , 2 . Simulation provides a controllable, efficient, and low-cost venue for both developing and testing autonomous vehicles (AV) 3 , 4 .…”

Section: Introductionmentioning

confidence: 99%

“…Simulation provides a controllable, efficient, and low-cost venue for both developing and testing autonomous vehicles (AV) 3 , 4 . But for simulation to be an effective tool, statistical realism of the simulated driving environment is a must 2 , 4 – 7 . In particular, the simulated environment needs to reproduce safety-critical encounters that AV might face in the real world with distribution-level accuracy.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Learning naturalistic driving environment with statistical realism

et al. 2023

Self Cite

View full text Add to dashboard Cite

For simulation to be an effective tool for the development and testing of autonomous vehicles, the simulator must be able to produce realistic safety-critical scenarios with distribution-level accuracy. However, due to the high dimensionality of real-world driving environments and the rarity of long-tail safety-critical events, how to achieve statistical realism in simulation is a long-standing problem. In this paper, we develop NeuralNDE, a deep learning-based framework to learn multi-agent interaction behavior from vehicle trajectory data, and propose a conflict critic model and a safety mapping network to refine the generation process of safety-critical events, following real-world occurring frequencies and patterns. The results show that NeuralNDE can achieve both accurate safety-critical driving statistics (e.g., crash rate/type/severity and near-miss statistics, etc.) and normal driving statistics (e.g., vehicle speed/distance/yielding behavior distributions, etc.), as demonstrated in the simulation of urban driving environments. To the best of our knowledge, this is the first time that a simulation model can reproduce the real-world driving environment with statistical realism, particularly for safety-critical situations.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Learning naturalistic driving environment with statistical realism

et al. 2023

Self Cite

View full text Add to dashboard Cite

show abstract

“…Inspired by the parallel and efficient processing of information in the biological brain, artificial neural networks (ANNs) have received a lot of attention and research − and have already achieved tremendous results in fields such as autonomous vehicles, biomedicine, natural language processing, and intelligent terminals . Most ANNs encode information as real-valued vectors for computation rather than as electrical spikes like the human brain, which leads to energy inefficiencies in ANNs.…”

Section: Introductionmentioning

confidence: 99%

Efficient Spiking Neural Networks with Biologically Similar Lithium-Ion Memristor Neurons

Ke,

Pan,

Jin

et al. 2024

ACS Appl. Mater. Interfaces

View full text Add to dashboard Cite

Benefiting from the brain-inspired event-driven feature and asynchronous sparse coding approach, spiking neural networks (SNNs) are becoming a potentially energy-efficient replacement for conventional artificial neural networks. However, neuromorphic devices used to construct SNNs persistently result in considerable energy consumption owing to the absence of sufficient biological parallels. Drawing inspiration from the transport nature of Na + and K + in synapses, here, a Li-based memristor (Li x AlO y ) was proposed to emulate the biological synapse, leveraging the similarity of Li as a homologous main group element to Na and K. The Li-based memristor exhibits ∼8 ns ultrafast operating speed, 1.91 and 0.72 linearity conductance modulation, and reproducible switching behavior, enabled by lithium vacancies forming a conductive filament mechanism. Moreover, these memristors are capable of simulating fundamental behaviors of a biological synapse, including long-term potentiation and long-term depression behaviors. Most importantly, a thresholdtunable leaky integrate-and-fire (TT-LIF) neuron is built using Li x AlO y memristors, successfully integrating synaptic signals from both temporal and spatial levels and achieving an optimal threshold of SNNs. A computationally efficient TT-LIF-based SNN algorithm is also implemented for image recognition schemes, featuring a high recognition rate of 90.1% and an ultralow firing rate of 0.335%, which is 4 times lower than those of other memristor-based SNNs. Our studies reveal the ion dynamics mechanism of the Li x AlO y memristor and confirm its potential in rapid switching and the construction of SNNs.

show abstract

A Survey on Testbench‐Based Vehicle‐in‐the‐Loop Simulation Testing for Autonomous Vehicles: Architecture, Principle, and Equipment

Cheng,

Wang,

Zhao

et al. 2024

Advanced Intelligent Systems

View full text Add to dashboard Cite

Autonomous vehicles (AVs) must be thoroughly tested to ensure safety and reliability before marketing. Simulation‐based testing has gained widespread recognition as the essential approach for AV testing by providing sufficient testing scenarios in the virtual environment. Vehicle‐in‐the‐loop (VIL) simulation has the ability to perform comprehensive tests and validations for the AVs’ overall behaviors while keeping significant testing accuracy and efficiency through the combination of the virtual scenarios and the physical AV. This article provides an overview of representative studies on testbench‐based VIL simulation testing for AVs, mainly focusing on utilizing testbenches to simulate realistic road conditions, and using physical signal stimulation methods and related equipment to generate sensors’ physical signals. This article first summarizes current AV testing studies, identifying existing issues and flaws of the state‐of‐the‐art methods and tools. Afterward, the testbench‐based VIL is addressed around architecture, principles, advantages, and characteristics. Then, the road condition simulation and the sensor physical signal generation in VIL are discussed in depth from structure, principle, and corresponding advanced equipment. Finally, research gaps between cutting‐edge technologies and AV testing applications in industrialization are identified to facilitate future research in this direction.

show abstract

Dense reinforcement learning for safety validation of autonomous vehicles

Cited by 171 publications

References 37 publications

Learning naturalistic driving environment with statistical realism

Learning naturalistic driving environment with statistical realism

Efficient Spiking Neural Networks with Biologically Similar Lithium-Ion Memristor Neurons

A Survey on Testbench‐Based Vehicle‐in‐the‐Loop Simulation Testing for Autonomous Vehicles: Architecture, Principle, and Equipment

Contact Info

Product

Resources

About