How Many Ways to Crash? TRB Annual Meeting
Nicolas Saunier
[email protected]
January 13th 2010
Collaboration with Karim Ismail, Clark Lim and Tarek Sayed University of British Columbia
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Outline
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results using Video Data
Conclusion
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
Road Safety Analysis
• Limits of the traditional approach based on historical
collision data: • Problems of availability and quality, • Insufficient data to understand the mechanisms that lead to
collisions, • Reactive approach.
• Need for proactive approaches and surrogate safety
measures that do not depend on the occurrence of collisions.
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
Surrogate Safety Measures
• Research on surrogate safety measures that • bring complementary information, • are related to traffic events that are more frequent than collisions and can be observed in the field, • are correlated to collisions, logically and statistically. • A traffic conflict is “an observational situation in which two
or more road users approach each other in space and time to such an extent that a collision is imminent if their movements remain unchanged” ´ 1977]. [Amundsen and Hyden,
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
The Safety/Severity Hierarchy
Accidents F I PD
Serious Conflicts Slight Conflicts Potential Conflicts Undisturbed passages
Various severity measures.
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
The Collision Course
• A traffic conflict is “an observational situation in which two
or more road users approach each other in space and time to such an extent that a collision is imminent if their movements remain unchanged”. • For two interacting road users, many chains of events may
lead to a collision. • It is possible to estimate the probability of collision if one
can predict the road users’ future positions.
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
What about Evasive Actions?
• Necessary by construction for traffic conflicts. • The severity of a traffic conflict does not depend on the
characteristics of the evasive action (e.g. the Swedish traffic conflict technique). • The emphasis on evasive actions is most likely related to
the traffic conflict collection techniques: emergency evasive actions are relatively easy to identify by observers in the field. • Future work: understand why collisions are avoided, and
the link between interactions with and without a collision.
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
Movement Prediction
• Learn road users’ motion patterns (including frequencies),
represented by actual trajectories called prototypes • Match observed trajectories to prototypes and extrapolate
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
A Simple Example
t1
0.7 0.3
t2 0.4
1
0.6
2
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
Collision Points
Using of a finite set of extrapolation hypotheses, enumerate the collision points CPn . Severity indicators can then be computed: X P(Collision(Ui , Uj )) = P(Collision(CPn )) Pn TTC(Ui , Uj , t0 ) =
P(Collision(CPn )) tn P(Collision(Ui , Uj ))
n
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
Motion Pattern Learning
Traffic Conflict Dataset, Vancouver 58 prototype trajectories (2941 trajectories)
Reggio Calabria, Italy 58 prototype trajectories (138009 trajectoires)
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Road User Tracking
Experimental Results
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Motion Prediction
Experimental Results
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Motion Prediction
Experimental Results
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Motion Prediction
Experimental Results
Conclusion
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Collision Probability
The Severity Indicators 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00
2
3
4 Time (second)
5
6
2
3
4 Time (second)
5
6
3.0 TTC (second)
Motivation
2.5 2.0 1.5 1.0 0.5
Parallel conflict, Kentucky dataset
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Road User Tracking
Experimental Results
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Motion Prediction
Experimental Results
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Motion Prediction
Experimental Results
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Motion Prediction
Experimental Results
Conclusion
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
The Severity Indicators
Collision Probability
0.5
TTC (second)
Motivation
0.4 0.3 0.2 0.1 0.0 3.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 3.0
3.5
4.0
4.5 5.0 Time (second)
5.5
6.0
3.5
4.0
4.5 5.0 Time (second)
5.5
6.0
Parallel collision, Kentucky dataset
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
Distribution of Severity Indicators
Histogram of Before-and-After TTC
12 TTC Before TTC After
500
Frequency of traffic events
Frequency of traffic events
600
400 300 200 100 0
0
2
4
6
8 TTC
min
10 12 in seconds
14
16
18
10
8
6
4
2
20
Histogram of Before-and-After PET quency of traffic events
Before2500 and after study, Oakland, CA. 2000 1500 1000
5 PET Before PET After
quency of traffic events
3000
4
3
2
1
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
Distribution of Severity Indicators (2)
Maximum Collision Probability 140 120 100 80 60 40 20 00.1 70 60 50 40 30 20 10 00.1
Traffic Conflicts
0.2
0.3
0.4
0.5 0.6 0.7 Collision Probability
0.8
0.9
1.0
60 50 40 30 20 10 00.0
1.0
30 25 20 15 10 5 00.0
Collisions
0.2
0.3
0.4
0.5 0.6 0.7 Collision Probability
Kentucky dataset.
0.8
0.9
Minimum TTC Traffic Conflicts
0.5
1.0
1.5
2.0 2.5 TTC (second)
3.0
3.5
4.0
4.5
2.0 2.5 TTC (second)
3.0
3.5
4.0
4.5
Collisions
0.5
1.0
1.5
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Spatial Distribution of the Collision Points Collisions 48 42 36 30 24 18 12 6 0
Kentucky dataset.
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Spatial Distribution of the Collision Points Traffic Conflicts 72 64 56 48 40 32 24 16 8 0
Kentucky dataset.
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
Conclusion
• Tools and framework for automated road safety analysis
using video sensors • Data mining and visualization for safety analysis • Future work: • Validation of proactive methods for road safety analysis (Clark Lim and Tarek Sayed at UBC) • Understanding and modelling of the mechanisms that lead ´ ´ to accidents (Ecole Polytechnique de Montreal) • Need for more open science: data and code sharing
http://nicolas.saunier.confins.net
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Questions ?
Experimental Results
Conclusion
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
´ C., editors (1977). Amundsen, F. and Hyden, Proceedings of the first workshop on traffic conflicts, Oslo, Norway. Institute of Transport Economics. Ismail, K., Sayed, T., and Saunier, N. (2010). Automated analysis of pedestrian-vehicle conflicts: A context for before-and-after studies. In Transportation Research Board Annual Meeting Compendium of Papers, Washington, D.C. 10-3739. Under consideration for publication in Transportation Research Record: Journal of the Transportation Research Board. Saunier, N. and Sayed, T. (2008). A Probabilistic Framework for Automated Analysis of Exposure to Road Collisions. Transportation Research Record: Journal of the Transportation Research Board, 2083:96–104. Saunier, N., Sayed, T., and Ismail, K. (2010).
Motivation
Probabilistic Framework for Automated Road Safety Analysis
Experimental Results
Conclusion
Large scale automated analysis of vehicle interactions and collisions. In Transportation Research Board Annual Meeting Compendium of Papers, Washington, D.C. 10-4059. Under consideration for publication in Transportation Research Record: Journal of the Transportation Research Board. Saunier, N., Sayed, T., and Lim, C. (2007). Probabilistic Collision Prediction for Vision-Based Automated Road Safety Analysis. In The 10th International IEEE Conference on Intelligent Transportation Systems, pages 872–878, Seattle. IEEE. Sonnenburg, S., Braun, M. L., Ong, C. S., Bengio, S., Bottou, L., Holmes, G., LeCun, Y., Muller, K.-R., Pereira, F., ¨ ¨ ¨ Rasmussen, C. E., Ratsch, G., Scholkopf, B., Smola, A., Vincent, P., Weston, J., and Williamson, R. (2007). The need for open source software in machine learning. Journal on Machine Learning Research, 8:2443–2466.
