PDT: Several examples of on-line measurement of driver cognitive workload

Transcription

1 PDT: Several examples of on-line measurement of driver cognitive workload Richard van der Horst & Marieke Martens

2 The problem: workload and distraction ITS the solution? 2

3 Behavioral Research by: Accident analysis Real-world observations/ Field Operational Tests Instrumented vehicles Driving simulator driver vehicle 3 Laboratory experiments Mathematical modelling + valid - - effects interpretable +

4 TNO driving simulator facilities 4

5 Behavioural measures Performance indicators: speed, speed variance, headway, TTC lateral placement, SDlat, TLC Visual attention/distraction Glance duration, number of glances Workload PDT, Tocc, physiological measures, subjective ratings Comfort vertical accelerations, subjective ratings Acceptance questionnaires, surveys 5

6 Car driving Continuous task (lane keeping) Planned actions (overtaking, turning) Sudden actions (braking lead vehicle) Changes in road environment (work zone) Use of in-vehicle system (navigation) How to measure workload while conducting these tasks? 6

7 Workload Secondary task - CMT (Continuous Memory Task) (mental workload) - Numerical 99 task (visual) - Peripheral Detection Task (PDT) Physiological measures - Heart rate frequency (overall level of activation) - Heart rate variability (cognitive or mental workload) - EEG (Electro Encephalo Gram) (P300 measure for attention) RSME (Rating Scale Mental Effort) One-dimensional. For driving as sensitive as multi-dimensional scales (e.g. NASA-TLX) 7

8 Workload measures Disadvantage of several measures: Measurement for longer periods of time/whole run Need for a measure: Suitable for short periods of time (peak loads) Able to measure variations in workload Objective Is not distracting attention Suitable for different workload conditions, due to: road environment traffic situation in-vehicle system etc. 8

9 Secondary tasks Secondary task performance depends on workload level primary task Verwey and Veltman (1995): secondary visual task sensitive to variations in task demand 9

10 Functional field of view decreases with increasing workload Miura (1986): with increasing traffic density, detection times for light spots on the windscreen increased. Visual field of view reduces with higher driving task complexity Williams (1982, 1985, 1995): higher reaction times to more peripheral stimuli with increasing workload Visual tunneling or cognitive tunneling? Van Winsum, Martens & Herland (1999): Peripheral Detection Task PDT 10

11 Peripheral Detection Task PDT First used in European project IN-ARTE A small red square presented on simulator screen for 1 s (not of relevance for primary task, easy to neglect) On average, each 4 s (random between 3-5 s) In an area degrees left and 2-4 degrees above horizon Response via finger switch (< 2s) Low-level easy-to-automate process that needs little conscious attention Sensitive to short lasting peaks in workload 11

12 Examples of PDT/workload studies Driving simulator studies IN-ARTE Roadwise WoW Road environment complexity On-the-road studies (instrumented vehicles) Type of display control Safety effects of navigation systems 12

13 Example 1: IN-ARTE study: Rural road Reaction time Fraction of missed signals 13

14 Example 1: IN-ARTE study: Motorway Reaction time Fraction of missed signals 14

15 Example 1: IN-ARTE study: type of support 15

16 Example 1: IN-ARTE study: visual or cognitive tunneling? Difference high/low workload No difference in horizontal angle -> cognitive tunneling 16

17 Example 2: WoW Driving simulator study Dutch DOT pilot FOT Wijzer op Weg : Roadwise Which information is suitable for direct presentation in the car? Possibilities for traffic management? Implications for traffic safety? Main Question for driving simulator study: Safe to do so on public road? Standing still: points of interest menu audio: news headlines speed warning 17

18 Example 2: WoW Driving behaviour speed headway Steering behaviour Lane keeping Time-To-Collision Workload questionnaire (RSME) Peripheral Detection Task (PDT) Scenarios: normal driving DRIP-info Sharp curve Rural roads City driving Work Zone Traffic jam by incident extreme effort very great effort great effort considerable effort rather much effort some effort a little effort almost no effort absolutely no effort Acceptance Standard questionnaire open question 18

19 Example 2: WoW Results PDT RSME: from >42.8 from just below to just above some effort 19

20 Example 3: Complexity of road environment and workload How to measure effects of motorway environments? A10 Beltway Amsterdam 20

21 Example 3: Complexity of road environment and workload 21

22 Example 3: Road environment complexity and workload 22

23 Example 4: Display Control Dashboard versus steering column Steering column: blind operation (after practice) Dashboard: buttons ask for eye-hand coordination Execution of task needs check on display Differences between dashboard and steering column with respect to visual attention and workload? 23

24 Example 4: Display control and workload Peripheral Detection Task (PDT) Secondary task Functional visual field of view reduces with increasing workload of driving task -> decreasing performance Rating Scale Mental Effort (RSME, BSMI in Dutch) Subjective measure Scale between

25 Example 4: Display control Results workload PDT (1) 25

26 Example 4: Display control and workload (2) RSME 26

27 Example 5: Safety effects of navigation systems Are navigation systems reducing workload? PDT 27

28 Example 5: Are navigation systems reducing workload? free waypoint navigation PDT reaction time PDT %missed signals subjective RSME 28

29 Conclusions PDT examples (1) IN-ARTE study PDT sensitive for variations in workload PDT makes distinction in workload: by task complexity (type of road, other traffic) By in-vehicle support modality WoW study Overall, workload increased (both RSME and PDT) But PDT could identify that: Workload increased for normal scenarios (free + car -following) For more critical scenarios increase in workload due to system diminished 29

30 Conclusions PDT examples (2) Road environment complexity simulator study: Subjective workload higher in the maximum version (including work zone), but around the level some effort (RSME) Experienced workload was larger for numerical task than for the PDT Overall, secondary tasks showed no workload effects (subjects are able to neglect distraction by the visual complex environment) Both secondary tasks sensitive to identify small workload effects for specific road sections (sharp curve and work zone) Display control on-the-road study: Both steering column and dashboard control resulted in a substantial increase of workload (both RSME and PDT) No effects on workload between type of control based on PDT Reduction of subjective workload (minus 20%) Navigation system on-the-road study: RSME: With navigation system workload rated much lower than with conventional aids PDT: % missed stimuli 20% lower; reaction time decreased with 12% 30

31 Overall conclusions PDT PDT is a visual secondary task but measuring cognitive workload based on cognitive tunneling paradigm PDT low-level easy-to-automate process that needs little conscious attention PDT suitable for measuring short-lasting peaks in workload PDT sensitive for variations in workload PDT makes distinction in workload: by task complexity (type of road, other traffic, critical scenarios) By in-vehicle support modality Head-mounted LED better than dashboard related Reaction time and % missed signals give similar results, but last one a bit more sensitive Criterion values can be derived based on normal and safety-critical scenarios (both relatively and absolutely) Ready for standardization 31

32 Literature Hoedemaeker, M., Hogema, J.H. & J.J.A. Pauwelussen (2006). Driving simulator study on the effects of road environment complexity on work load. (TNO report TNO-DV C244). Soesterberg: TNO Human Factors (in Dutch). Hogema, J.H. (2005). Roadwise driving simulator study. (TNO report TNO-DV C008). Soesterberg: TNO Human Factors (in Dutch). Martens, M.H. & W. van Winsum (2000). Measuring distraction: the Peripheral Detection Task. Proceedings NHTSA Internet Forum on the safety impact of driver distraction when using in-vehicles technology. Washington, D.C.: NHTSA. Miura, T. (1986). Coping with situational demands: a study of eye movements and peripheral vision performance. In: A.G. Gale, I.D. Brown, C.M. Haslegrave, P. Smith & S. Taylor (eds.), Proceedings of Vision in Vehicles. Amsterdam: Elsevier, North Holland, Verwey, W.B. & J.A. Veltman (1995). Measuring workload peaks while driving. A comparison of nine common workload assessment techniques. TNO report TNO-TM 1995 B04). Soesterberg: TNO Human Factors. Vonk, T., Rooijen, T. van, Hogema, J. & P. Feenstra (2007). Do navigation systems improve traffic safety? (TNO report 2007-D-R0048/B). Delft: TNO Mobility and Logistics. Williams, L.J. (1995). Peripheral target recognition and visual field narrowing in aviators and nonaviators. The International Journal of Aviation Psychology 5, Williams, J.L. (1985). Tunnel vision induced by a foveal load manipulation. Human Factors, 27, Williams, L.J. (1982). Cognitive load and the functional field of view. Human Factors, 24, Winsum, W. van & M. Hoedemaeker (2000). A road test of a prototype satellite system for in-vehicle menu control. (TNO report TM-00-C003). Soesterberg: TNO Human Factors. Winsum, W. van, Martens, M. & L. Herland (1999). The effects of speech versus tactile driver support messages on workload, driver behaviour and user acceptance. (TNO report TM-99-C043). Soesterberg: TNO Human Factors. 32