Safety assessment of highly automated driving systems in test tracks: A new framework

Feng, Shuo; Feng, Yiheng; Yan, Xintao; Shen, Shengyin; Xu, Shaobing; Liu, Henry

doi:10.1016/j.aap.2020.105664

Cited by 54 publications

(32 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…where q(x) is called the importance function. By introducing importance functions, the testing priority of critical scenarios will be improved, so does the evaluation efficiency [14][15][16][17] . However, all existing IS-based methods suffer from the "curse of dimensionality" 19 , and thus cannot be applied directly for the complex driving environment.…”

Section: Theoremmentioning

confidence: 99%

“…Towards solving the inefficiency issue, scenario-based approaches have been proposed. Based on the importance sampling (IS) theory, critical scenarios can be purposely designed for accelerating the efficiency of AV evaluation [12][13][14][15][16][17] . However, existing scenario generation methods can only be applied for scenarios that involve simple maneuvers of a very limited number of vehicles with very short duration, for instance, a cut-in maneuver from a background vehicle for a few seconds.…”

mentioning

confidence: 99%

See 1 more Smart Citation

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

et al. 2021

Self Cite

View full text Add to dashboard Cite

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.

show abstract

Section: Theoremmentioning

confidence: 99%

mentioning

confidence: 99%

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

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…We need a dataset containing a massive collection of road images and steering angles against that image for a deep learning predictive model. Different nations’ legislators (e.g., the USA, China, Australia, Singapore, and South Korea) [ 62 , 63 , 64 , 65 , 66 , 67 , 68 ] have established or are adopting different regulating measures to enhance the security and privacy of data utilized and sent by autonomous cars. The gathering of data on public roadways is essential for self-driving car autonomy [ 69 ].…”

Section: Methodology and Implementationmentioning

confidence: 99%

Delicar: A Smart Deep Learning Based Self Driving Product Delivery Car in Perspective of Bangladesh

Chy

Masum

Sayeed

et al. 2021

Sensors

View full text Add to dashboard Cite

The rapid expansion of a country’s economy is highly dependent on timely product distribution, which is hampered by terrible traffic congestion. Additional staff are also required to follow the delivery vehicle while it transports documents or records to another destination. This study proposes Delicar, a self-driving product delivery vehicle that can drive the vehicle on the road and report the current geographical location to the authority in real-time through a map. The equipped camera module captures the road image and transfers it to the computer via socket server programming. The raspberry pi sends the camera image and waits for the steering angle value. The image is fed to the pre-trained deep learning model that predicts the steering angle regarding that situation. Then the steering angle value is passed to the raspberry pi that directs the L298 motor driver which direction the wheel should follow. Based upon this direction, L298 decides either forward or left or right or backwards movement. The 3-cell 12V LiPo battery handles the power supply to the raspberry pi and L298 motor driver. A buck converter regulates a 5V 3A power supply to the raspberry pi to be working. Nvidia CNN architecture has been followed, containing nine layers including five convolution layers and three dense layers to develop the steering angle predictive model. Geoip2 (a python library) retrieves the longitude and latitude from the equipped system’s IP address to report the live geographical position to the authorities. After that, Folium is used to depict the geographical location. Moreover, the system’s infrastructure is far too low-cost and easy to install.

show abstract

“…For example, Zhao et al (36,37) introduced importance sampling techniques and generated testing cases for car-following and lane-changing maneuvers. To solve the overvalue problem of worst cases, Feng et al (38)(39)(40)(41) proposed a new definition of scenario criticality, which can be computed as a combination of maneuver challenge and exposure frequency, and generated critical cases using optimization methods and reinforcement learning techniques on various environment settings, including cut-in scenarios and car-following scenarios. To further address the challenge brought by the high dimensionality of complex environments (e.g., highway driving), Feng et al (12) proposed a framework of generating a naturalistic and adversarial driving environment by adding sparse but adversarial adjustments to the NDE.…”

Section: Corner Case Generation For Vehicle Decision-makingmentioning

confidence: 99%

“…In this paper, the problem formulation is consistent with (12,(38)(39)(40)(41). Let u describe the pre-determined parameters of the operational design domains (ODDs), such as the number of lanes, weather, and so forth.…”

Section: Problem Formulationmentioning

confidence: 99%

Corner Case Generation and Analysis for Safety Assessment of Autonomous Vehicles

Sun

Feng

Yan

et al. 2021

Transportation Research Record: Journal of the Transportation R

Self Cite

View full text Add to dashboard Cite

Testing and evaluation is a crucial step in the development and deployment of connected and automated vehicles (CAVs). To comprehensively evaluate the performance of CAVs, it is necessary to test the CAVs in safety-critical scenarios, which rarely happen in a naturalistic driving environment. Therefore, how to purposely and systematically generate these corner cases becomes an important problem. Most existing studies focus on generating adversarial examples for perception systems of CAVs, whereas limited efforts have been put into decision-making systems, which is the highlight of this paper. As the CAVs need to interact with numerous background vehicles (BVs) for a long duration, variables that define the corner cases are usually high-dimensional, which makes the generation a challenging problem. In this paper, a unified framework is proposed to generate corner cases for decision-making systems. To address the challenge brought by high dimensionality, the driving environment is formulated based on the Markov decision process, and the deep reinforcement learning techniques are applied to learn the behavior policy of BVs. With the learned policy, BVs behave and interact with the CAVs more aggressively, resulting in more corner cases. To further analyze the generated corner cases, the techniques of feature extraction and clustering are utilized. By selecting representative cases of each cluster and outliers, the valuable corner cases can be identified from all generated corner cases. Simulation results of a highway driving environment show that the proposed methods can effectively generate and identify the valuable corner cases.

show abstract

Safety assessment of highly automated driving systems in test tracks: A new framework

Cited by 54 publications

References 8 publications

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

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

Delicar: A Smart Deep Learning Based Self Driving Product Delivery Car in Perspective of Bangladesh

Corner Case Generation and Analysis for Safety Assessment of Autonomous Vehicles

Contact Info

Product

Resources

About