Understanding and Predicting Human Driving Behaviors via Machine Learning Models (Multiple Year Project)

Principal InvestigatorYi-Ching Lee, PhD, The Children's Hospital of Philadelphia; Co-PI: Santiago Ontanon, Drexel University

 

Below is an executive summary of this line of research. Please note that each summary describes results and interpretation that may not be final. Final interpretation of results will be in the peer-reviewed literature.

 

Among teen drivers, speed management is a significant factor that contributes to crashes. To better understand how technology can help reduce crash risk, this study used driving simulators to analyze and learn a human behavior – speed management – using a technique called machine learning. Such techniques are used to create predictive driving models that may be incorporated into smart-car or in-vehicle monitoring systems. The long-term goal of this research was to fully understand young drivers’ risky behaviors in order to advance safety awareness and management of driving risks.

 

The driving simulator used as part of the Simulated Driving Assessment.

Year 1: 2013-2014

In phase one of the study, researchers determined the feasibility of using computational approaches to automatically model speed management. The project utilized a subset of an existing dataset from the development of a Simulated Driving Assessment. This Assessment included driving scenarios and situations that reflected the most common crash configurations among young drivers as identified in the results of the National Motor Vehicle Crash Causation Survey (NMVCCS). The dataset contained behavioral and eye-tracking data from four simulator drives, a driver education instructor’s subjective ratings of driving behaviors, images from three video cameras, and self-report measures from the teen participants.

Data from 17 novice teen drivers (licensed for less than 90 days) were analyzed using three modeling techniques. The modeling techniques predicted the behavior of the driver at the same instant, as represented by a given sample, and the future behavior of the driver. Additionally, a certified driver education instructor reviewed video recordings of the participant’s simulator drives and evaluated each participant’s crash likelihood and driving skill levels in comparison to all drivers, as well as teenage drivers.

Results suggest that modeling was effective at predicting vehicle control behaviors up to one second into the future. These predictions were consistent with the skill proficiency and crash likelihood ratings given by a professional driving education instructor.

 

Year 2: 2014-2015

Drawing on phase one results, the objectives of phase two were to determine: (1) which parameters of modeling techniques were most suitable for producing predictions of unsafe and dangerous speed management behaviors, and (2) the effect of feedback on behavioral changes in speed management in subsequent simulator drives.

Researchers upgraded the machine learning modeling techniques, including testing and comparing machine learning approaches for accuracy and scalability; improving model accuracy; reducing prediction errors; and redesigning the feedback interface. In parallel, researchers developed and tested driving scenarios in the driving simulator to best align with the collection of training and testing scenarios for the machine learning modeling work. The driving contexts were programmed in a specific way and presented in a specific order so that the machine learning models could capture participants’ driving styles in various driving contexts. This ensured that the models had sufficient training data with which to output predictions on how each driver would perform in new driving scenarios.

The third component of the project involved collecting data from young drivers. Drivers in the feedback condition received model predictions as feedback half-way through a study visit. Drivers in the control condition did not receive any feedback. This allowed researchers to examine the effect of receiving feedback on subsequent driving behaviors in similar or different driving contexts.

Preliminary results show that the indirect model outperforms all other modeling approaches. This work paves the way for future in-vehicle machine learning systems that can provide drivers with real-time monitoring and feedback to help reduce crash risk.

This indirect model can be used to predict accurate long-term driving behaviors. The figure shows 
four predictions, in four different situations. For each prediction, two trajectories are shown: The blue dots
represent the trajectory that the driver actually took, while the red dots show the predicted trajectory by our model. 

  

Year 3: 2015-2016

Machine learning is an analytic technique that shows promise in providing algorithms to help predict and manage certain teen driver behaviors that can contribute to crash risk, including speed management. In this multi-year study, researchers aimed to create a real-time feedback system to help teen drivers improve their speed management skills.

In Phase One of the study, researchers determined the feasibility of using machine learning-based computational approaches to automatically model speed management in a driving simulator. Their findings suggest that modeling was effective at predicting vehicle control behaviors up to one second in the future. These predictions matched the skill proficiency and crash likelihood ratings given by a professional driving education instructor. In Phase Two of the study, researchers evaluated the effectiveness of providing customized feedback from these machine learning models to young drivers and found that one of the models tested can be used to accurately predict long-term driving behaviors.

While promising, these results were not obtained in real-time but rather through off-line processing of completed drives. Therefore, Phase Three of the study aimed to make this technology closer to real life by exploring machine learning models that could make predictions in real time, as the teen participants were driving in the simulator. The objectives were to: (1) develop new machine learning algorithms for updating models and generating predictions in real-time, (2) update the data collection infrastructure to allow real-time transfer of driving simulator data, (3) design effective feedback mechanisms to generate predictions and examine the effect of such feedback on driver behaviors, and (4) explore the feasibility of incorporating additional data sources, such as physiological measurements and eye-tracking data.

The study team successfully created a mechanism to receive real-time data from the driving simulator, to train machine learning models using this data, to make predictions of participants' driving behavior, and to provide feedback about speed management to participants while driving in the simulator. The four drives used in this project contained a variety of road configurations, terrains, road conditions, and speed limit zones. While data analysis is ongoing, this work will help to inform the development of interventions to help teen drivers develop crucial speed management skills, a major factor contributing to crashes, and shows great potential in preventing other dangerous driving behaviors via personalized feedback models.

The researchers also envision a future where real-time feedback via machine learning techniques will be part of in-vehicle assistance systems that can track driver performance and provide meaningful, adaptive guidance to novice drivers to enhance safety.

The simulator lab, housing a Pontiac G6 driver seat, LCD panels allowing for a 160 degree field of view,
active pedals and steering system, and a rich audio environment.

 

 

 

Project Team Members:
Chelsea Ward, MS, The Children’s Hospital of Philadelphia (Y3); Flaura Winston, MD, PhD, The Children’s Hospital of Philadelphia (Y1,Y2); Catherine McDonald, PhD, RN, The University of Pennsylvania (Y1); Avelino Gonzalez, PhD, University of Central Florida (Y1,Y2); Dana Bonfiglio, BS, The Children’s Hospital of Philadelphia (Y1).

Students:
Molly Tiedeken, Drexel University (Y1,Y2, Y3); Sam Snodgrass, Drexel University (Y1,Y2,Y3); Leif Malm, Drexel University (Y3);Celia Cheng, Drexel University (Y2). 

IAB Mentors:
Doug Longhitano, American Honda Motor Co., Inc. (Y2,Y3); Melissa Miles, State Farm Mutual Automobile Insurance Company (Y3); Tim Hsu, FCA US LLC (Y2); Noelle Lavoie, Parallel Consulting (Y1); Steve Roberson, State Farm Mutual Automobile Insurance Company (Y1); Dan Glaser, General Motors Holdings LLC (Y1); Clayne Woodbury, Realtime Technologies Inc. (Y1); Sara Seifert, Minnesota HealthSolutions (Y1).

 

About This Center

This Center is made possible through a grant from the National Science Foundation (NSF) which unites CHOP, University of Pennsylvania, and The Ohio State University researchers with R&D leaders in the automotive and insurance industries to translate research findings into tangible innovations in safety technology and public education programs.

FIND US ON: SafetyMD Twitter CChIPS LinkedIn

You are here: Home Portfolio CChIPS 2014-2015 Research Portfolio Understanding and Predicting Human Driving Behaviors via Machine Learning Models (Multiple Year Project)