DriveLab: A New Tool for Driving Simulator Research

April 24, 2014

Driving simulators offer a safe, highly reproducible environment for assessing driver behavior. However, reducing the data to easy-to-interpret metrics can be extremely time-consuming and effortful. Even worse, it can be error-prone. My recent research involves the development of a tool to help standardize driving simulator results called DriveLab.

DriveLab is a set of widely applicable routines for reducing simulator data to expert-approved metrics. It was developed along with the Simulated Driving Assessment (SDA), a set of driving exercises that expose teen drivers to real-world crash scenarios. These potential crash scenarios replicate in the driving simulator environment the most common causes of fatal crashes as identified by the National Motor Vehicle Crash Causation Survey (NMVCCS) conducted by the National Highway Traffic Safety Administration (NHTSA). DriveLab was developed around Matlab™, which is widely used by researchers both in academic and industrial settings. After being validated in two research studies at CHOP, the newly created Matlab™ toolbox will be generalized for use in all driving simulators.

Interestingly, one automobile manufacturer told me at the end of my talk at the recent SAE World Congress & Exhibition in Detroit that his company generates a lot of data on its driving simulator and is confronted daily with the data reduction process. By providing an efficient and flexible context for data reduction, visualization, and analysis, DriveLab can potentially be used by the scientific community and industry to systemize and harmonize driver simulator analyses. No longer will driving simulator researchers have to rely on different definitions for driving metrics, making it difficult to compare results.

Controlled scenarios in the driving environment, such as traffic jams, pedestrians, time of day, and weather, can be easily programmed in a simulator. Teens' awareness of crash risks and management of those risks is of great interest to all of us that are working to develop interventions to prevent teen crashes, the number 1 killer of young adults. A driving simulator allows us to study how teens react to critical situations by examining whether they recognize potential risks and how they control the vehicle. More research is needed to develop simulator protocols that can reliably assess driving performance under conditions that typically result in crashes. For instance, we are now studying failure-to-brake errors so that we can add relevant scenarios to our SDA’s set of driving exercises.