Dataset Information

Intelligent driving intelligence test for autonomous vehicles with naturalistic and adversarial environment.

ABSTRACT: Driving intelligence tests are critical to the development and deployment of autonomous vehicles. The prevailing approach tests autonomous vehicles in life-like simulations of the naturalistic driving environment. However, due to the high dimensionality of the environment and the rareness of safety-critical events, hundreds of millions of miles would be required to demonstrate the safety performance of autonomous vehicles, which is severely inefficient. We discover that sparse but adversarial adjustments to the naturalistic driving environment, resulting in the naturalistic and adversarial driving environment, can significantly reduce the required test miles without loss of evaluation unbiasedness. By training the background vehicles to learn when to execute what adversarial maneuver, the proposed environment becomes an intelligent environment for driving intelligence testing. We demonstrate the effectiveness of the proposed environment in a highway-driving simulation. Comparing with the naturalistic driving environment, the proposed environment can accelerate the evaluation process by multiple orders of magnitude.

SUBMITTER: Feng S

PROVIDER: S-EPMC7854639 | biostudies-literature | 2021 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Intelligent driving intelligence test for autonomous vehicles with naturalistic and adversarial environment.

Feng Shuo S Yan Xintao X Sun Haowei H Feng Yiheng Y Liu Henry X HX

Nature communications 20210202 1

Driving intelligence tests are critical to the development and deployment of autonomous vehicles. The prevailing approach tests autonomous vehicles in life-like simulations of the naturalistic driving environment. However, due to the high dimensionality of the environment and the rareness of safety-critical events, hundreds of millions of miles would be required to demonstrate the safety performance of autonomous vehicles, which is severely inefficient. We discover that sparse but adversarial ad ...[more]

PMID: 33531506

Similar Datasets

Project description:The development of science and technology continues to promote the progress of society. The current intelligence and automation technology has become widely used in society. To this end, this study proposes a vehicle intelligent control system based on edge computing and deep learning to promote the far-reaching development of intelligent technology and automation technology. First, control algorithms are used to design a switch control strategy combining accelerator and brake. Second, a fuzzy control algorithm based on vehicle tracking and trajectory deviation is designed to enhance the vehicle's stability during steering. A Convolutional Neural Network (CNN) is used to recognize the car's surroundings as it drives. In addition, accelerator and brake controllers and vehicle tracking and trajectory deviation controllers are connected to the vehicle's wiring. Then, the data transmission function based on edge computing is applied to the vehicle's intelligent control system. Finally, trajectory tracking and emergency braking experiments are carried out on the control system to verify the practicability and reliability of the method and the effectiveness of CNN. The simulation experiments are carried out on two states of medium speed and high speed to verify the effectiveness of the longitudinal anti-collision system of the test vehicle when the target vehicle suddenly decelerates. The results demonstrate that the driving speed of the experimental vehicle is set to 50km/h, the distance between the experimental vehicle and the target vehicle is 40m, and the target vehicle in front drives at a constant speed of 50km/h. The target vehicle in front of the car suddenly decelerates in 5 seconds, and the speed drops to 0 after 5 seconds. The actual distance between the experimental vehicle and the target vehicle is very close to the expected safe space, and the experimental vehicle is in a safe state during this process. When the experimental vehicle starts to decelerate, the experimental vehicle adopts emergency deceleration to ensure a safe distance between the two vehicles. At this time, the car enters the second-level early warning state, but driving safety can still be guaranteed. It is advisable to maintain low-speed emergency braking in this state. This study provides creative research ideas for the follow-up research on the intelligent control system of uncrewed vehicles and contributes to the development of intelligence and automation technology.

Dataset Information

Intelligent driving intelligence test for autonomous vehicles with naturalistic and adversarial environment.

Publications

Intelligent driving intelligence test for autonomous vehicles with naturalistic and adversarial environment.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets