Bicycle Naturalistic Data Collection

Final Report

Proposal Excerpt

When considering driving safety, there are four major factors that contribute to vehicle crashes: vehicle design, infrastructure design, driver abilities and behavior, and legal determinants. These factors are exacerbated when observing smaller and lighter vehicles such as motorcycles and bicycles. When observing motorcycle and bicycle vehicle design, three elements create the most danger to the rider: 1) the unstable two-wheel in-line alignment, 2) the lack of an exterior safety shield, and 3) vehicle size and weight, particular as compared to larger four or more wheeled vehicles. Infrastructure design is mostly static; therefore, the onus is on the driver to navigate their vehicle safely when facing dynamic problems like changes in lighting setting, traffic patterns, and weather conditions. Also, infrastructure does not make accommodation for vehicle purpose and size regardless of navigation or destination.

Dangerous situations also occur from driver limited abilities, attention, and behavior. the majority of navigation and adaptation come from how well a driver is operating their vehicle, and how quickly drivers can react when immediate changes occur, like slick roads, or a quickly stopping vehicle in front of them. Since it is difficult to determine driver mental acuity, research must rely on observing how drivers react and behave, and since this is the most unpredictable and dynamic of the four factors, it is therefore important to rely on naturalistic driver research to determine what behaviors lead to car crashes. The final danger determination on the highway is legal factors, which is a broader scope of human factors looking at drivers collectively instead of individually. Dangerous situations can occur from common flouting of legal rules (like not stopping at a red light), or in a gap in the need for safety laws (like lack of helmet laws for motorcyclists and bicyclists that would increase safety).

Considering the danger that bicyclists are faced with, preventing and mitigating crashes involving these groups of people is the focus of this research effort. When analyzing the factors that led to crashes involving motorcycles/bicycles and how they combine and relate to each other, it was evident that connected vehicle infrastructure (CVI) technology can be utilized to mitigate a significant number of these danger factors. Bicycle locations can be broadcasted by on-board devices from bicycles that can be received by other equipped vehicles to avoid possible conflicts with these vehicles and issue warnings to the drivers and/or infrastructure.

For whatever reasons (e.g. inattention, distraction, etc.), drivers and cyclists clearly fail to obey traffic rules at both signalized and sign-controlled intersections. Hence the problem is how to prevent/mitigate these intersection-related crashes that involves bicycles. A naturalistic experiment will be designed to collect realistic bicycle behavior to study the cyclist behavior when approaching and crossing intersections. The research objectives are:

1) Collect and analyze bicycle naturalistic data for different riders to assess their behavior when approaching and crossing intersections,
2) Assess the applicability of the collected bicycle naturalistic data to develop models to predict if a cyclist is going to violate a red light,
3) Assess the applicability of the collected bicycle naturalistic data to develop models to predict if a cyclist is going to violate a stop sign
4) Collect naturalistic bicycle data that can be used in future project

Highlights

The following results were obtained using only a part of the data. The data collection is still ongoing and the results will be updated after completing the analysis for the entire data.
A naturalistic cycling data collection system was conducted that can be used to develop bicycle violation prediction models.
The system architecture that embodied cyclist violation prediction models was demonstrated. It was shown how connected vehicle technology can be adopted for different parts to communicate amongst themselves. Communication between different system entities was shown to have different purposes:

1. Sending required variables for violation prediction models such as bicycle speed, acceleration, and current location.
2. Sending warning to the drivers/riders in potential danger.
3. Sending a “control change” order to change the signal setting.

In case of signal-controlled intersections, it was found that it is more likely that a cyclist violates a red light when making right turns. Also, the probability of red light violation decreases when there is side traffic at the intersection or when there is traffic in front of the cyclist.
In case of stop-controlled intersections, the likelihood of violating a stop sign increases when there is no side traffic or when the cyclist is younger.
Violation prediction models at stop-controlled intersections were developed. The model accuracy of about 94% was obtained when the time to intersection for the cyclist was about 2 seconds.

Publications

Jahangiri A., Elhenawy M., Rakha H., and Dingus T. (2016). Investigating Cyclist Violations at Signal-Controlled Intersections Using Naturalistic Cycling Data. 19th IEEE Intelligent Transportation Systems Conference, Rio de Janeiro, Brazil, Nov. 1-4.

Jahangiri, A.,M. Elhenawy, H. Rakha, and T. A. Dingus. Studying cyclist violations at stop sign-controlled intersections using naturalistic cycling data. Published within the proceedings of the 95th Transportation Research Board Annual Meeting, 2016.

Arash Jahangiri, Hesham A. Rakha, Thomas A. Dingus, Developing a System Architecture for Cyclist Violation Prediction Models Incorporating Naturalistic Cycling Data, Procedia Manufacturing, Volume 3, 2015, Pages 5543-5550, ISSN 2351-9789, https://dx.doi.org/10.1016/j.promfg.2015.07.724.

Jahangiri, A., Rakha, H. (2015). “Applying Machine Learning Techniques to Transportation Mode Recognition using Mobile Phone Sensor Data,” IEEE Transactions on Intelligent Transportation Systems, vol.PP, no.99, pp.1,12.

A. Jahangiri, H. Rakha, T. A. Dingus (2015). “Predicting Red-light Running Violations at Signalized Intersections using Machine Learning Techniques”. Published within the proceedings of the 94th Transportation Research Board Annual Meeting, Washington, DC, January 2015 (No. 15-2910).

A. Jahangiri, H. Rakha (2015). “Distributed Learning: An Application to Transportation Mode Identification”. Published within the proceedings of the 94th Transportation Research Board Annual Meeting, Washington, DC, January 2015.

Jahangiri A. and Rakha H. (2014), “Developing a Support Vector Machine Classifier for Transportation Mode Identification by Using Mobile Phone Sensor Data,” Published within the proceedings of the 93rd Transportation Research Board Annual Meeting, Washington DC, January 12-16, CD-ROM [Paper # 14-1442].

Presentations

Jahangiri A., Elhenawy M., Rakha H. and Dingus T. (2017). Studying Cyclist Violatons at Stop Sign-Controlled Intersections using Naturalistic Cycling Data. 96th Transportation Research Board Annual Meeting, Washington DC, January 7-12. [Paper # 17-00214]

Jahangiri A., Elhenawy M., Rakha H., and Dingus T. (2016), “Investigating Cyclist Violations at Signal-Controlled Intersections Using Naturalistic Cycling Data,” 19th IEEE Intelligent Transportation Systems Conference, Rio de Janeiro, Brazil, Nov. 1-4.

A. Jahangiri, H. Rakha, and T. Dingus (2015), “Developing a System Architecture for Cyclist Violation Prediction Models Incorporating Naturalistic Cycling Data,” 6th International Conference on Applied Human Factors and Ergonomics (AHFE), Las Vegas, USA July 26-30.

A. Jahangiri, H. Rakha, T. A. Dingus (2015). “Predicting Red-light Running Violations at Signalized Intersections using Machine Learning Techniques”. Presented at the 94th Transportation Research Board Annual Meeting, Washington, DC, January 2015 (No. 15-2910).

A. Jahangiri, H. Rakha (2015). “Distributed Learning: An Application to Transportation Mode Identification”. Presented at the 94th Transportation Research Board Annual Meeting, Washington, DC, January 2015.

Jahangiri A. and Rakha H., (2015), “Transportation Mode Detection using a Distributed Learning Approach,” 22nd ITS World Congress, Bordeaux, France, Oct. 5-9.

A. Jahangiri, H. Rakha (2014). “Transportation Mode Recognition using Smartphone Sensor Data”. Presented at DriveSense’14: NSF Workshop on Large-Scale Traffic and Driving Activity Data, Old Dominion University, Norfolk, VA, October 2014.

A. Jahangiri, H. Rakha (2014). “Utilization of Smartphone Sensor Data to Develop Detection and Prevention Models in Transportation”. Presented at the 4th Annual CEE Student Research Day Poster Session, Virginia Tech, Blacksburg, VA, 2014.

Jahangiri A. and Rakha H. (2014), “Developing a Support Vector Machine Classifier for Transportation Mode Identification by Using Mobile Phone Sensor Data,” Presented at the 93rd Transportation Research Board Annual Meeting, Washington DC, January 12-16, CD-ROM [Paper # 14-1442].

Sponsor Organization

Research and Innovative Technology Administration
University Transportation Centers Program
Department of Transportation
1200 New Jersey Avenue, SE
Washington, DC 20590
USA

UTC Grant Manager

Harwood, Leslie
Phone: 540-231-9530
Email: lharwood@vtti.vt.edu

Performing Organization

Virginia Polytechnic Institute and State University, Blacksburg
Virginia Tech Transportation Institute
3500 Transportation Research Plaza
Blacksburg, Virginia 24061
USA

Research Investigators

Dingus, Tom
Jahangiri, Arash
Rakha, Hesham

More Information

RiP URL
Project Poster
TriD Format