Driving Research: Advances in Methods

Abstract:

At a high level, the goal of this project is to collect data on how drivers approach and drive through intersections. Research at the University of Michigan concerns the development of the driving simulator scenarios and testing human subjects. In addition, considerable related research is being conducted on research methods. At MIT, models based on the research at Michigan are being developed. This abstract concerns those part of the project conducted at the University of Michigan.

Key outcome 1: Augmented reality warnings can reduce crashes. An important element of this research was examining how drivers responded to warnings that were presented on the windshield, what is sometimes referred to as a full-windshield head-up display. Three driving simulator experiments were conducted in which subjects drove though a series of 70 intersections multiple times. We found there were significant performance advantages for augmented reality warnings, now being documented in several technical reports.

Key outcome 2: Honking leads to subjects driving more realistically. Most driving simulators used are fixed base. They do not move. Other cues may be missing as well. As a consequence, drivers may not drive in a realistic manner. One example of such is that when distracted or loaded, people drive more slowly than normal. To overcome that problem, we had a simulated vehicle follow the subject, honking if the subject drove more than 5 mi/hr below the speed limit mid-block, or delayed in accelerating after a traffic light changed. We found, depending on the rule, subjects drove 5-8 mi/hr faster, driving at the desired speed. Furthermore, the software changes to achieve this outcome is minimal, 5 more lines of code per intersection.

Key outcome 3: SAE Recommended Practice J2944 was published in June. This is probably the most important publication the PI has written. It was written because driving performance measures and statistics have not been commonly named (we found 13 names for lane departure) or defined (about 10-15%) of the time, making it almost impossible to compare driving studies. This 171-page, 300+ reference document defines about 50 terms including brake response time (about 15 versions), road departure (4 versions), time to collision (2 versions), lane departure (11 versions), and so forth. The next step is to get journals and conferences to require compliance with this practice or papers will be rejected.

Key outcome 4: Improved driving simulators have been developed. We have developed and then upgraded the simulator used for this project in the Department of Industrial and Operations Engineering. That simulator is now being used for follow on research funded by Nissan and Hyundai. In addition, we developed a process for providing low-cost driving simulators for use by undergraduate (and graduate) students. That involves providing access to and software for the OpenDS free driving simulator, and by treating the steering wheel/pedal assemblies as reserve books, making them available to students at the library. This approach overcomes the security and support cost issues that limit access to research facilities.