A Causal Model of Intersection-Related Collisions for Drivers With and Without Visual Field Loss

Biebl, Bianca; Kacianka, Severin; Unni, Anirudh; Trende, Alexander; Rieger, Jochem W.; Lüdtke, Andreas; Pretschner, Alexander; Bengler, Klaus

doi:10.1007/978-3-030-90966-6_16

Cited by 4 publications

(6 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Bareinboim et al [10] proposed a causal inference-based method to solve data fusion problems in the context of big data. Biebl et al [14] presented a causal model to predict accident risks in an intersection for drivers with impairments. In addition to works focusing on AI [11,18,19,59], some works have applied causality to the security analysis of CPSs [34,36,37,41,42].…”

Section: Related Workmentioning

confidence: 99%

ACAV: A Framework for Automatic Causality Analysis in Autonomous Vehicle Accident Recordings

Sun,

Poskitt,

Sun

et al. 2024

Proceedings of the IEEE/ACM 46th International Conference on Software Engineering

View full text Add to dashboard Cite

The rapid progress of autonomous vehicles (AVs) has brought the prospect of a driverless future closer than ever. Recent fatalities, however, have emphasized the importance of safety validation through large-scale testing. Multiple approaches achieve this fully automatically using high-fidelity simulators, i.e., by generating diverse driving scenarios and evaluating autonomous driving systems (ADSs) against different test oracles. While effective at finding violations, these approaches do not identify the decisions and actions that caused them-information that is critical for improving the safety of ADSs. To address this challenge, we propose ACAV, an automated framework designed to conduct causality analyses for AV accident recordings in two stages. First, we apply feature extraction schemas based on the messages exchanged between ADS modules, and use a weighted voting method to discard frames of the recording unrelated to the accident. Second, we use safety specifications to identify safety-critical frames and deduce causal events by applying CAT-our causal analysis tool-to a station-time graph. We evaluated ACAV on the Apollo ADS, finding that it can identify five distinct types of causal events in 93.64% of 110 accident recordings generated by an AV testing engine. We further evaluated ACAV on 1206 accident recordings collected from versions of Apollo injected with specific faults, finding that it can correctly identify causal events in 96.44% of the accidents triggered by prediction errors, and 85.73% of the accidents triggered by planning errors. CCS CONCEPTS• Computer systems organization → Embedded and cyberphysical systems; Maintainability and maintenance; • Software and its engineering;

show abstract

Section: Related Workmentioning

confidence: 99%

ACAV: A Framework for Automatic Causality Analysis in Autonomous Vehicle Accident Recordings

Sun,

Poskitt,

Sun

et al. 2024

Proceedings of the IEEE/ACM 46th International Conference on Software Engineering

View full text Add to dashboard Cite

show abstract

“…RQ2: Building on our premise that accountability can be expressed with SCMs, we can rely on work on how to assess, measure and compare SCMs. Here we use work by [Halpern, 2016, Chapter 4] who gives different examples and guidelines on how to structure models, and our own work Pretschner, 2021] on how to identify and leverage accountability patterns in SCMs.…”

Section: Rq1mentioning

confidence: 99%

“…Since the SCM of a system derives from the system design, we can also reverse that process and design a system so that it conforms to a specific SCM. We show how the accountability patterns from RQ2 can be used to assess the accountability model of a system and then adapt the system design such that causes of effects are easy to identify Pretschner, 2021].…”

Section: Rq3mentioning

confidence: 99%

“…We need this to sensibly reason over the interaction between humans and machines, but unfortunately expressing human behavior formally is a much harder problem. In [Kacianka et al, 2019b] and later [Biebl et al, 2021], we made some first steps into this direction, but there is still too much that is not known. First, getting reliable models of general human behavior is hard and made worse by the fact that in a specific, or actual causal, scenario an individual human might stray very far from any norm.…”

Section: Future Workmentioning

confidence: 99%

See 1 more Smart Citation

Accountability for Cyber- Physical Systems

Pretschner

Kacianka

2021

Encyclopedia of Cryptography, Security and Privacy

Self Cite

View full text Add to dashboard Cite

First and foremost I would like to thank my mother Andrea and my sister Nike for their unwavering support throughout my whole life and especially during the writing of this thesis. Thank you for lifting me up when I was down, for pushing me when I was lazy and for proof-reading every word I ever wrote.Next and most importantly, I want to thank my supervisor Alexander Pretschner.Through this joint journey, you taught me how to get to the point, to tell a good story, and to be a decent researcher and human being. You had my back when I did mess up and supported me when I tried something new. I am especially grateful for the honest and tough discussions -they made this text and all my work so much better. Thank you for shaping my view of science, good work, and my philosophy of life.A special thanks goes to all my colleagues past and present. You made the chair something I was looking forward to almost every single day. We made so many wonderful memories, shared stories and insights. All of you shaped my thinking and attitude and I am a better man for it. I am proud to have worked with you and grateful to be able to consider you friends. Thank you.Last, but not least, I want to thank my friends for their help, support and especially for the time not spent on the thesis. Anastasia, for making this last leg of my PhD so special; Celia, Eva, Evi, Hannes, Luis, Marci, Maresa, Michael, Peer, and Philip for those countless hikes and climbs that kept me sane; Niina and Jan for our project and the discussions that inspired me so much; Anne, Corinna, Luisa, Markus, and Nikolaus for those long nights and deep talks.

show abstract

“…Studies have found affected persons to exhibit increased exploration times and unsystematic scanning patterns in visual search tasks (Schuett et al, 2009a ). In the context of driving, visual exploration is specifically challenging due to the high number of unnaturally fast-moving objects (Pomarjanschi et al, 2013 ), which results in reduced safety levels and increased crash involvement among some drivers with VFL (Johnson and Keltner, 1983 ; Haymes et al, 2008 ; Kwon et al, 2016 ; Biebl et al, 2021 ). While regarded as highly beneficial for driving safety, the ability to develop appropriate compensatory strategies shows high heterogeneity among individuals with VFL (Bowers et al, 2009 ; Elgin et al, 2010 ; Hardiess et al, 2010 ; Wood et al, 2011 ; Papageorgiou et al, 2012 ; Haan et al, 2014 ; Bahnemann et al, 2015 ; Kübler et al, 2015 ; Alberti et al, 2017 ).…”

Section: Introductionmentioning

confidence: 99%

Opportunities and Limitations of a Gaze-Contingent Display to Simulate Visual Field Loss in Driving Simulator Studies

Biebl

Arcidiacono

Kacianka

et al. 2022

Front. Neuroergonomics

Self Cite

View full text Add to dashboard Cite

BackgroundResearch on task performance under visual field loss is often limited due to small and heterogenous samples. Simulations of visual impairments hold the potential to account for many of those challenges. Digitally altered pictures, glasses, and contact lenses with partial occlusions have been used in the past. One of the most promising methods is the use of a gaze-contingent display that occludes parts of the visual field according to the current gaze position. In this study, the gaze-contingent paradigm was implemented in a static driving simulator to simulate visual field loss and to evaluate parallels in the resulting driving and gaze behavior in comparison to patients.MethodsThe sample comprised 15 participants without visual impairment. All the subjects performed three drives: with full vision, simulated left-sided homonymous hemianopia, and simulated right-sided homonymous hemianopia, respectively. During each drive, the participants drove through an urban environment where they had to maneuver through intersections by crossing straight ahead, turning left, and turning right.ResultsThe subjects reported reduced safety and increased workload levels during simulated visual field loss, which was reflected in reduced lane position stability and greater absence of large gaze movements. Initial compensatory strategies could be found concerning a dislocated gaze position and a distorted fixation ratio toward the blind side, which was more pronounced for right-sided visual field loss. During left-sided visual field loss, the participants showed a smaller horizontal range of gaze positions, longer fixation durations, and smaller saccadic amplitudes compared to right-sided homonymous hemianopia and, more distinctively, compared to normal vision.ConclusionThe results largely mirror reports from driving and visual search tasks under simulated and pathological homonymous hemianopia concerning driving and scanning challenges, initially adopted compensatory strategies, and driving safety. This supports the notion that gaze-contingent displays can be a useful addendum to driving simulator research with visual impairments if the results are interpreted considering methodological limitations and inherent differences to the pathological impairment.

show abstract

A Causal Model of Intersection-Related Collisions for Drivers With and Without Visual Field Loss

Cited by 4 publications

References 24 publications

ACAV: A Framework for Automatic Causality Analysis in Autonomous Vehicle Accident Recordings

ACAV: A Framework for Automatic Causality Analysis in Autonomous Vehicle Accident Recordings

Accountability for Cyber- Physical Systems

Opportunities and Limitations of a Gaze-Contingent Display to Simulate Visual Field Loss in Driving Simulator Studies

Contact Info

Product

Resources

About