, PLS Regression): PLS REGRESSION, WIREs Comp Stat, vol.2, pp.97-106
, Gaze behavior during take-595 over after a long period of autonomous driving: A pilot study, Proc. Int. Conf. Driving 596 Simulation Conference Europe VR, 2018.
, , 2019.
, Long automated driving phase affects take-over performance. IET Intelligent Transport 599 Systems
, Mind wandering during everyday 601 driving: An on-road study, Accident Analysis & Prevention, vol.122, pp.76-84, 2019.
, Control task 604 substitution in semiautomated driving: Does it matter what aspects are automated?, Human 605 Factors, vol.54, pp.747-761, 2012.
, Advancements, prospects, and impacts of automated driving systems, 2017.
, International Journal of Transportation Science and Technology, vol.6, pp.208-216
, Toward a theory of situation awareness in dynamic systems, Human, vol.610, pp.32-64, 1995.
, The Out-of-the-Loop Performance Problem and 612 Level of Control in Automation, Human Factors, vol.37, pp.381-394, 1995.
, Driving performance after self-regulated control 615 transitions in highly automated vehicles, Human Factors, vol.59, pp.1233-1248, 2017.
, How the duration of 618 automated driving influences take-over performance and gaze behavior, Advances in, p.619, 2017.
, Ergonomic Design of Systems, Products and Processes, pp.309-318
, Distracted driver performance to 622 multiple alerts in a multiple-conflict scenario, Human Factors, vol.56, 2014.
, Partially automated driving as a 625 fallback level of high automation, 6. Tagung Fahrerassistenzsysteme. Der Weg zum 626 automatischen Fahren. (TÜV SÜD Akademie GmbH), 2013.
, Using markov chains to 628 understand the sequence of drivers' gaze transitions during lane-changes in automated 629 driving, Proceedings of the International Driving Symposium on Human Factors in Driver, p.630, 2019.
,
, , 2019.
, Detection and prediction of driver drowsiness using artificial neural network models, Accident, vol.640
, Analysis & Prevention, vol.126, pp.95-104
, Vigilance decrement and 642 passive fatigue caused by monotony in automated driving, 6th International Conference on 643 Applied Human Factors and Ergonomics, pp.2403-2409, 2015.
, Driver Inattention During 645 Vehicle Automation: How Does Driver Engagement Affect Resumption Of Control?, 4th 646 International Conference on Driver Distraction and Inattention, pp.1-16, 2015.
, Were they in the loop during 648 automated driving? Links between visual attention and crash potential, Injury Prevention, vol.23, pp.649-281, 2016.
, A Methodology for Inducing the Out of the Loop Phenomenon 651 in Highly Automated Driving, International Conference on Traffic and Transport 652 Psychology, 2016.
, Engaging with Highly Automated Driving: To be 654 or Not to be in the Loop, Proceedings of the Eighth International Driving Symposium on, 2015.
, Human Factors in Driver Assessment, Training and Vehicle Design, vol.656, pp.190-196
, Are you in the loop? Using gaze dispersion to understand 658 driver visual attention during vehicle automation, Transportation Research Part C: Emerging 659 Technologies, vol.76, pp.35-50, 2017.
, How much time do drivers need to obtain 661 situation awareness? A laboratory-based study of automated driving, Applied Ergonomics, vol.60, pp.662-293, 2017.
, Eye movements and hazard perception in active 664 and passive driving, Visual cognition, vol.23, pp.736-757, 2015.
, Driver adaptation to haptic shared control of the 666 steering wheel, 2014 IEEE International Conference on Systems, Man, and Cybernetics, vol.667, pp.1505-1509, 2014.
, Where we look when we drive with or without active 669 steering wheel control, PLoS One, vol.7, p.43858, 2012.
, Highly Automated Driving, p.671, 2012.
, Secondary Task Performance, and Driver State, Human Factors, vol.54
, The 674 "out-of-the-loop" concept in automated driving: Proposed definition, measures and 675 implications, Technology & Work, vol.21, pp.87-98, 2019.
, Getting, 2019.
URL : https://hal.archives-ouvertes.fr/hal-02108005
, Back Into the Loop: The Perceptual-Motor Determinants of Successful Transitions out of 678
, Automated Driving, Human Factors, vol.61, pp.1037-1065
, Monitoring an automated system for a single failure: 680 Vigilance and task complexity effects, Human Factors, vol.38, pp.311-322, 1996.
, Obstacle avoidance under automated steering: 683 Impact on driving and gaze behaviours, Transportation Research Part F: Traffic Psychology 684 and Behaviour, vol.43, pp.315-324, 2016.
, Fatigue and voluntary 686 utilization of automation in simulated driving, Human Factors, vol.54, pp.734-746, 2012.
, Humans and Automation: Use, Misuse, Disuse, p.689, 1997.
, Human Factors, vol.39, pp.230-253
, Taxonomy and Definitions for Terms Related to On-Road Motor, 2016.
, Vehicle Automated Driving Systems
, Active 693 and Passive Fatigue in Simulated Driving: Discriminating Styles of Workload Regulation and 694 Their Safety Impacts, J Exp Psychol Appl, vol.19, pp.287-300, 2013.
, Looking at the Road 696 When Driving Around Bends: Influence of Vehicle Automation and Speed, Front. Psychol, vol.697, p.10, 2019.
, The information that drivers use: is it indeed 90% visual?, Perception, vol.25, pp.1081-699, 1996.
, Human error taxonomies applied to driving: A 701 generic driver error taxonomy and its implications for intelligent transport systems, Safety 702 Science, vol.47, pp.227-237, 2009.
, Blink rate: a possible measure of fatigue, 1994.
, Human Factors, vol.36, pp.285-297
, R: A language and environment for statistical computing, 2013.
, Keeping eye and mind on the road, 2005.
, Uppsala Dissertations from the Faculty of Social Sciences, p.83
, Sensitivity of eye-movement measures to 709 in-vehicle task difficulty, Transportation Research Part F: Traffic Psychology and Behaviour, vol.710, issue.8, pp.167-190, 2005.
, Asleep at the automated 712 wheel-Sleepiness and fatigue during highly automated driving, Accident Analysis & 713 Prevention, vol.126, pp.70-84, 2019.
, The pls Package: Principal Component and Partial 715, 2007.
, Least Squares Regression in R, Journal of Statistical Software, vol.18, pp.1-24
, Research on 717 vehicle-based driver status/performance monitoring; development, validation, and refinement 718 of algorithms for detection of driver drowsiness, National Highway Traffic Safety, vol.719, 1994.
, Administration Final Report
, What determines the take-over time? An 721 integrated model approach of driver take-over after automated driving, Accident Analysis & 722 Prevention, vol.78, pp.212-221, 2015.
, Why is steering not the same as 724 braking? The impact of non-driving related tasks on lateral and longitudinal driver interventions during conditionally automated driving, Transportation Research Part F: 726 Traffic Psychology and Behaviour, vol.50, pp.65-79, 2017.
, Appendix 1: Step-by-step procedure of PLS, vol.9
, This appendix develops step-by-step the procedure to predict the MW score (Y) using PLS regressions
, We used a training dataset (matrix of gaze behaviour computed over a given time window t, Xt) and a 732 validation dataset (matrix of gaze behaviour computed over the two final minutes of automated driving, p.733
,
, Steps A and B were performed with the training datasets and enabled computing the optimal parameters 735 (number of components and relevant visual indicators) of the prediction models. With that configuration, 736 the accuracy of the model was tested for both the training and validation datasets
, Step A: Calculating the optimal number of components
, With many components, the model will be complex and highly accurate but also 743 very specific of the data. By contrast, few components will mean a simpler model structure. The model 744 may lose accuracy but may be more generalizable to other datasets, PLS models are based on several orthogonal components, which constitute the underlying structure of 742 the prediction model
, This compromise was sought by testing several numbers of components (from one to 10 components)
, The optimal number of components would minimize the mean square error of prediction (MSEP), with Step B: Reducing the number of visual indicators 754 In the previous step, an optimal structure of the prediction model was found, considering all possible 755 visual indicators (182) to predict the MW score
, Because PLS regression is a linear statistical model, the relationship between the training dataset (Xt) 761 and the dependent variable to estimate ? t
, Coefficients can be interpreted as follows: 766 -Coefficient signs indicate the direction in which a visual indicator (from Xt) influenced the 767 estimation of the MW score (? t, train ). If positive, the MW score increased
, A coefficient's magnitude (absolute value) indicates the importance of each indicator relative
, If the magnitude of a coefficient was close to zero, the contribution of this visual indicator to 771 the prediction would be negligible
, To reduce the number of visual indicators of Xt, coefficient magnitudes were compared with an 777 increasing threshold value, which ranged from 0.01 to 0.2. A new PLS regression was computed for 778 each partial matrix (i.e. a matrix comprising only the indicators whose coefficient magnitude, p.779