0
Upcoming SlideShare
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Standard text messaging rates apply

# Session 48 Holger Rootzén

296

Published on

Published in: Technology, Automotive
0 Likes
Statistics
Notes
• Full Name
Comment goes here.

Are you sure you want to Yes No
• Be the first to comment

• Be the first to like this

Views
Total Views
296
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
1
0
Likes
0
Embeds 0
No embeds

No notes for slide

### Transcript

• 1. Statistiska metoder kring kritiska händelser och olyckor Holger Rootzén, Mathematical Sciences, Chalmers http://www.math.chalmers.se/~rootzen/
• 2. How can information from near-crashes be used to prevent real crashes? 1 Do near-crashes resemble real crashes? Are more extreme near-crashes more like real crashes? 2 Is it possible to find driver behavior or traffic situations which is different in near-crashes than in normal driving? Are these differences even more extreme in real crashes? Statistical methods
• 3. 1Regression: Is relative risk the same for crashes and fornear-crashes?Extreme Value Statistics (EVS): Can the distribution ofnear-crashes predict the frequency of real crashes?Extreme value statistics models behavior in extremesituations. It is a well developed area of statistics whichincludes all the standard statistical tools: suitabledistributions (the Extreme Value distributions), estimationmethods, model checking tools, multivariate and regressiontools, user friendly (free) software
• 4. Crash proximity measures:TTEC = Time To edge CrossingGap = time between first car leaves conflict area and second car enters conflict areaTTC = Time To Collisionetc
• 5. UMTRI (Gordon et al) “do near-crashes give similarrisk estimates as crashes?”Seemingly Unrelated Regression  yes to question 1 (?)EVS: TTEC  road departure  road way departure crash2.3 mile segment of US-23 with117 traversals by 43 differentdrivers in instrumented cars.EV distribution fit to minimum TTEC values for the 117traversals  predicts 12 road departures/yearOn the average there were 1.8 road way departurecrashes/year yes to question 1 (?)
• 6. Slide from presentation by P. Tarko, Purdue university: ”Riskfour-year actual counts of daytime right-angle collisions evaluation for intersections” 50 model-based estimates of crash frequency □Notes: year actual counts of daytime rightangle collisions four (1) Model-based crash frequency estimate of site 97903 is EVS estimate of crash frequency from gap measurements ● excluded from this plot. The value of the estimate is 387. 40 (2) Error bars represent 95% Poisson confidence intervals Error bars show 95% Poisson confidence intervals based on observedDaytime right-angle collisions based on the observed counts. counts 30 8-hour observations of crossing gaps. Signalized intersections in the Lafayette area. Summer 2003 20  yes to question 1 (?) 10 0 site site site site site site site site site site site site 87905 87906 87907 87909 87915 87930 87933 97901 97903 97905 97911 97920
• 7. We (J. Jonasson, R. Jörnsten, O. Nerman) “do near-crashes givesimilar risk estimates as crashes?”100-car data, risk of rear-ending, TTC EVS estimate of expected number of crashes ≈ 0.5 Observed number of crashes = 10 Didn’t work?
• 8. An explanationAll real rear-ending crashes were in start-stop traffic but the near-crashes with usable TTC were in higher speed situationsSo maybe still:  yes to question 1 (?)
• 9. The next steps: question 2• Use near-crashes for investigation of to what extent attention measures and other driving and traffic characteristics influence crash risk• Develop statistical predictors of crash risk• Investigate the relation of risk estimates obtained in different in naturalistic driving studies (Semifot, 100-car, SHRP 2, …)• Study the normal driving – near-crash/crash relation in naturalistic driving experimentsMore and better data crucial, new statistical methods must bedeveloped
• 10. M. Barnes*, A. Blankespoor, D. Blower, T. Gordon, P. Green, L. Kostyniuk, D. LeBlanc, S. Bogard.,B. R. Cannon, and S.B. McLaughlin (2010). Development of Analysis Methods Using RecentData: A Multivariate Analysis of Crash and Naturalistic Event Data in Relation to HighwayFactors Using the GIS Framework. Final Report SHRP S01, University of MichiganTransportation Research InstituteA. P. Tarko, P. Songchitruksa (2006). ESTIMATING FREQUENCY OF CRASHES AS EXTREMETRAFFIC EVENTS. Report, Purdue University