How do drivers avoid collisions? A driving simulator-based study

Li, Xiaomeng; Rakotonirainy, Andry; Yan, Xuedong

doi:10.1016/j.jsr.2019.05.002

Cited by 36 publications

(4 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…5 and in Table II To further investigate the safety/reliability performance of the proposed self-improvement approach, we have evaluated Generation-3 RL agents on the unseen safety-critical scenarios where τ ttc < 5s. A TTC threshold to resemble a risk of the scenario event is chosen to be τ ttc < 5s, as research in [16] considers τ ttc ∈ [3.5s, 5.0s] as a safetycritical condition. 1000 scenario events are collected by applying CE, BO, and AMS algorithms on all Generation-3 trained RL agents and selecting τ ttc < 5s samples randomly.…”

Section: Self-improvement Resultsmentioning

confidence: 99%

“…However, safety of these trained RL agents are not completely verified on the continuous range of rare-events where timeto-collision (TTC) measurements are low and probability of collision is high. In another conducted study [16], the effects of TTC on the high likelihood of collisions are analyzed with driver participants using the simulation environments.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Self-Improving Safety Performance of Reinforcement Learning Based Driving with Black-Box Verification Algorithms

Dagdanov¹,

Durmuş²,

Üre³

2022

Preprint

View full text Add to dashboard Cite

In this work, we propose a self-improving artificial intelligence system for enhancing the safety performance of reinforcement learning (RL) based autonomous driving (AD) agents based on black-box verification methods. RL methods have enjoyed popularity among AD applications in recent years. That being said, existing RL algorithms' performance strongly depends on the diversity of training scenarios. Lack of safetycritical scenarios in the training phase might lead to poor generalization performance in real-world driving applications. We propose a novel framework, where the weaknesses of the training set are explored via black-box verification methods. After the discovery of AD failure scenarios, the training of the RL agent is re-initiated to improve the performance of the previously unsafe scenarios. Simulation results show that the proposed approach efficiently discovers such safety failures in RL-based adaptive cruise control (ACC) applications and significantly reduces the number of vehicle collisions through iterative applications of our method.

show abstract

Section: Self-improvement Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Self-Improving Safety Performance of Reinforcement Learning Based Driving with Black-Box Verification Algorithms

Dagdanov¹,

Durmuş²,

Üre³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Helman and Reed [23], in their study "Validation of the Driver Behaviour Questionnaire Using Behavioural Data from an Instrumented Vehicle and High-Fidelity Driving Simulator" reported that driving simulation is relatively valid, however the use of the DBQ is a valid measure of driving behaviour. Li et al [24], in a driving-simulator-based study, reported that the use of cell phones by drivers tends to reduce driver brake reaction times. Xiaomeng et al also reported that the use of mobile phones in hands-free mode does not eliminate the safety problems associated with distracted driving, as it impairs driving performance in the same way as hand-held mobile phones.…”

Section: State Of the Artmentioning

confidence: 99%

Road User Attitudes and Their Reported Behaviours in Abuja, Nigeria

Uhegbu

Tight

2021

Sustainability

View full text Add to dashboard Cite

The continuous increases in the numbers of road traffic crashes (RTC) over the years, especially in developing countries, have been a source of worry. The majority of the RTC are attributed to road user behaviours exhibited by the drivers. This study sets out to investigate the road user attitudes and behaviours in Abuja, Nigeria. A total of 1526 questionnaires were distributed, of which 321 questionnaires were completed and returned. The questionnaires tried to understand four major road user behaviours, namely use of seatbelts, drink driving, use of mobile phone while driving, and use of child restraints. The results after analysing the questionnaires showed that the majority of the road users in Abuja showed high non-compliance with the use of seatbelts, either when driving or when being driven. About 64% of the respondents admitted to not using seatbelts. Results from the cross-tabulation process showed that the high non-compliance to seatbelt usage was statistically associated with young, single road users. Around one-third of Abuja road users admitted to drink driving. Road users who were married engaged in more frequent drink driving than road users who were single, and the association was statistically significant. A high percentage of Abuja road users admitted to using mobile phones while driving and 55.8% of the total respondents admitted to not using child restraints while driving. A lack of child restraints was statistically associated more with male road users than female road users. It is recommended that stricter enforcement of road safety laws should be undertaken and that the government should provide road safety agents with the right equipment (e.g., speed guns, breathalyzers) that would aid road safety agents to perform their duties effectively in order to curb the excessive bad road user behaviours in Abuja.

show abstract

“…In the simulators, the tests were performed a many tests e.g., to analyse the performance of the old drivers [1][2], drivers with various chronic diseases, e.g., Parkinson's or Alzheimer's disease [3], or [34 -37]. Threat factors affecting the behaviour of drivers [38][39], and possible risks occurring during the traffic [40][41][42], the impact of road shaping [43] in this intersection [44][45], may cause drivers' response times to change [46][47][48][49]. The response time may change when one has to be activated with time pressure [50].…”

Section: Introductionmentioning

confidence: 99%

The Impact of the Safe following Distance onto the Traffic Safety

Jurecki,

Poliak

2024

Communications

View full text Add to dashboard Cite

This paper provides an analysis of the values of safety distance between the two moving vehicles, based on the parameters of the behaviour of young drivers. The main objective of this research was to determine the driver's response time for braking manoeuvre (BRT) in car-following situations. The test results were used to verify certain recommendations for the vehicle driving parameters, the principles of driver's performance to increase road traffic safety, etc. The driver's response times, determined during the testing in the simulator, were used to assess the recommended values of the safe following distance under different road conditions and various ways of driver behaviour. In the paper, for different values of BRT a safety distance in emergency traffic situations was determined.

show abstract

How do drivers avoid collisions? A driving simulator-based study

Cited by 36 publications

References 24 publications

Self-Improving Safety Performance of Reinforcement Learning Based Driving with Black-Box Verification Algorithms

Self-Improving Safety Performance of Reinforcement Learning Based Driving with Black-Box Verification Algorithms

Road User Attitudes and Their Reported Behaviours in Abuja, Nigeria

The Impact of the Safe following Distance onto the Traffic Safety

Contact Info

Product

Resources

About