The National Advanced Driving Simulator (NADS)

OPTOMETRY Clinical Research

The National Advanced Driving Simulator (NADS) Description and Capabilities in Vision Related Research Mark Wilkinson, O.D., Timothy Brown, Ph.D., and Omar Ahmad, M.S. KEYWORDS Driving simulator, contact lens, intraocular lens, visual field loss

Abstract This article reviews the advantages of performing driving research in a high fidelity driving simulator where the driver can be challenged in ways that could lead to property damage and personal injury if done on a test track or public roadway. Several vision related studies performed at the University of Iowa National Advanced Driving Simulator are review in this article to demonstrate the capabilities of these devices.

NADS Background The National Advanced Driving Simulator (NADS), located at the University of Iowa, is a research facility that specializes in conducting driving simulation research for government and industry sponsors. The facility contains several driving simulators including the NADS-1 owned by the National Highway Traffic Safety Administration, which is the largest, highest fidelity, and most advanced ground vehicle research simulator in the United States. The facility also houses the NADS-2, a static base simulator, and the MiniSim™, a portable desktop simulator. Driving simulators enable researchers to simulate real-world driving situations that would be too risky or complicated to reliably re-create on the road or on a test track. Driving simulation studies typically have a duration of 3-12 months and have a shorter time duration than naturalistic studies. Most relevant to medical research, driving simulators serve as the only reliable tool for re-creating lighting, traffic, and road conditions over a large cohort of test subjects.

hardware components that comprise the physical simulator. This makes it possible to conceptualize studies on one simulator and then transfer the experiment to other simulators with minimum effort. It also allows for multiple aspects of vision related driving to be evaluated while minimizing the time and cost of development. The NADS-1 driving simulator consists of an entire car, sport utility vehicle, or truck cab located inside a 24-foot dome (shown in Figure 1). Each vehicle cab is equipped with instrumentation specific to its make and model. All dashboard indicators are operational, and the control switches are instrumented to record driver inputs. Multiple in-vehicle cameras provide customized views of the cab environment including the driver and vehicle controls.

The NADS team of highly qualified researchers have conducted numerous simulator-based and on-the-road studies in the areas of driver performance measurement1 behavior and impairment,2 highway safety, collision avoidance, and invehicle technology,3 simulator-based training and assessment, instrumented vehicle use for driver model and simulator validation, and core simulator use and fidelity trade-off. Simulator Descriptions Driving simulators are described in terms of their fidelity, which is their ability to realistically immerse the driver into a driving experience in a virtual environment. The components that affect a simulator’s fidelity consist of both hardware and software. There are no published standards on how these components translate into fidelity. While the sensitivity of some simulators to specific protocols have been published,4, 5 no standardized protocols have been applied to a range of simulators to provide comparative assessments of the fidelity of various simulators. All simulators at NADS use the same core software. The differences in the simulators are mostly related to the

Figure 1 NADS 1 Dome The motion system consists of a 13 degree of freedom motion base. The dome sits on a turntable that turns +/- 330 degrees. This is mounted on a hexapod that produces roll, pitch, and yaw. The hexapod sits on a platform that moves laterally and longitudinally on a track area of over 4000 sq. feet. The vehicle cab is also mounted on high-frequency vibration actuators that

accurately produce what it feels like to drive different types of road surfaces including smooth, semi-smooth, rough and gravel. The visual imagery wraps 360 degrees around the driver and higher resolution is utilized in the forward field-of-view to accommodate better feature recognition and reduce eye fatigue. NADS-1 is accompanied with a rich set of virtual environments that depict urban, suburban, rural, and highway driving areas in day and nighttime driving conditions. The audio system generates sounds that emulate road, wind, tire, engine, and other vehicle noise, as well as special effects such as tire blowouts. Sample night-time driving environment are shown in Figure 2, 3 and 4.

Figure 2 Night scene urban

Figure 3 Night scene rural A sophisticated scenario control system is used to simulate other intelligent traffic within the virtual environment. A rich set of coordinators allows researchers to orchestrate events that reliably re-occur during the simulation. Almost every aspect of the virtual environment can be specified, including time of day, weather conditions, road coefficient of friction, vehicle interactions, and sudden mechanical failures. Some examples are yellow light dilemma, a system failure, the sudden appearance of a pedestrian/animal in the roadway, a skid on ice, or an oncoming vehicle drifting across the centerline. A digital video recording system allows the simultaneous recording of up to four streams of video. Approximately 15 independent video signals, including in-vehicle cameras, image generator outputs, motion bay cameras and in-dome overview

cameras, can be independently fed to any of the four recording streams. Streams are automatically saved on hard-disc, but can also be transferred to tape, DVDs or other media. The data acquisition system can record any of the internal simulator variables at a rate of 240 times a second, or any whole subdivision (i.e., 120, 60, 30, etc.). The data streams include information about the state of the driver’s vehicle (speed, acceleration, and position), surrounding traffic, eye tracking, head tracking, and driver posture. The suite of NADS simulators also includes the NADS-2, a fixed base simulator with a configurable field of view. For vision related studies, the NADS-2 can be configured with either plasma screens or high resolution projectors. These configurations can provide sufficient pixel density to provide a system that is not vision limited for drivers with a visual acuity of 20/20. This simulator also has a physical light source that is linked to the software to reproduce glare from the headlights of oncoming vehicles.6 To assess visual performance, a variety of measures can be calculated based upon interaction with the driving environment. Rudimentary measures such as the visual angle of objects when they are detected or identified can be used to determine the equivalent distance from the subject to the object. Safety relevant measures such as time-to-collision at points of time in the response can also be calculated to assess how visual performance effects safety and crash risk. 1,7-11 Finally, the NADS suite of simulators includes the NADS MiniSim™, a small footprint, and low cost alternative to the more realistic NADS-1 and NADS-2 simulators. This portable simulator is easily deployable at multiple sites and can be customized to meet a project’s specific needs. The MiniSim can be configured to have single or multiple displays, and runs on one PC. The system can be configured to have a high-quality steering wheel and pedals that can be mounted on a desk or built into a sophisticated cab to the user’s specifications. This simulator can also be configured to provide clear resolution for subjects with 20/20 vision. The MiniSim is based on the same software technology as the NADS-1 and NADS-2, ensuring compatibility of scenarios and data across simulators. Sample Vision Related Studies Contact Lens Pilot Study. Every year, new contact lenses are being developed with the goal of providing better vision to the recipients. Clinical trials take place to evaluate the clinical performance of these new lenses. Although these trials spend significant amounts of time assessing visual performance, they do so from a clinical rather than an applied perspective.  This study endeavored to determine how great the difference in driving performance is for individuals driving under mesopic conditions with standard spherical designed contact lenses versus aspheric designed contact lenses.

IOL (model LI61AO) and the AMO Tecnis aspheric silicone IOL (model Z9000). The primary endpoint for this analysis was the recognition distance for roadway obstacles. Secondary measures included detection distances for objects, as well as recognition and detection distances for roadway signs.

Figure 4 Night scene rural with deer Because spherical aberration increases as the pupils naturally dilate under lower light conditions causing a defocusing of the image on the retina, correction of spherical aberration was expected to increase retinal image clarity, thus allowing for a greater margin of safety when operating a motor vehicle under low light conditions. The study evaluation involved a rural drive at 55 mph during which time the subject was instructed to detect and identify sign content, and to detect and avoid road hazards. The primary analysis focused on data that could support parametric statistical analyses. A mixed model repeated measures technique was employed using the specified independent variables and their interactions. The independent variables included lens type (aspheric versus conventional lenses) and glare condition (present or absent). The dependent measures that were analyzed using parametric analyses included: angular size of objects when detected, reaction time to moving hazards, required deceleration needed to avoid moving hazards, response time to stationary hazards, peak lateral acceleration during avoidance for moving and stationary hazards, percentage of moving and stationary hazards hit, percentage of moving and stationary objects detected, recognition distance for signs, percentage of signs recognized correctly, and percentage of information on the signs identified correctly. Twenty-one of the forty participants enrolled in this pilot study completed all study procedures with sixteen failing the screening, two withdrawing consent and one dropped from the study. Results identified differences in performance between the two lenses tested for guide sign recognition, and the angular size of an incurring car when the driver responded. Drivers with the conventional lenses correctly recognized 95% (SE = 1.0%)of the green directional signs; whereas, drivers with the aspheric lenses recognized 92% (SE= 1.7%) of the signs (F=8.15, p = 0.0055). For angular size of the incurring car at driver response, drivers with the aspheric lenses responded when the car was at an angular size of 0.11 radians (SE = 0.03); whereas, drivers with the conventional lenses did not respond until the angular size was 0.16 radian (SE = 0.01). This represents a quicker response (F=3.93, p = 0.005) to the incurring car of 950 feet for drivers wearing the aspheric lenses, versus the conventional lenses. These results indicate that that conventional lenses may provide a benefit for reading certain types of road signs, whereas, aspheric lenses may aid drivers in identifying changes in vehicle behavior earlier allowing drivers to respond sooner. Intraocular Lens Study. Another study was designed to identify and quantify differences in nighttime driving performance between the Bausch & Lomb SofPort Advanced Optics aspheric silicone

No statistical main effect differences were found between the two lenses. When examining the interactive effects, there was only one interaction that was statistically significant between the two IOL types: detection distances for roadway obstacles. This analysis yielded a three-way interaction between lens type, time of day, and roadway (F=5.98, p=0.0184). When examining the threeway interaction amongst group, roadway, and time of day, there was no meaningful difference in object detection distance between lens groups in any of the time/roadway configurations with the possible exception of the urban/dusk configuration, where the Tecnis Z9000 lens (M=961, SE = 23) resulted in a detection distance of approximately 80 feet greater than the SofPort LI61AO lens (M=879, SE = 24). However, a simple effects t-test on this difference found that it was not statistically significant (p>0.05). When taking into account the speed and detection distance, the improvement, even had it been statistically significant, would have had little meaningful practical impact when considered in terms of preview time. For the Tecnis Z9000 lens, drivers would have averaged approximately 13 seconds to identify and respond to the object; whereas for the SofPort LI61AO lens, drivers would have had approximately 12 seconds. With both sets of lenses, drivers had sufficient preview time to detect, identify, and respond to the signs and objects in the virtual environment. Both lenses meet the needs of drivers for driving at dusk and at night. Overall, this study showed that there were few differences between the Tecnis Z9000 lens and the SofPort LI61AO lens. The only statistical difference observed was a slightly better number of detections for the objects with the SofPort AO lens compared to the Tecnis Z9000 lens. Although statistically significant, the magnitude of this difference was not clinically meaningful. This study confirmed that the NADS simulator has the ability to detect significant differences, if present, as demonstrated by the large number of statistically significant effects for roadway type, time of day, and type of sign or object. The sensitivity of the study is demonstrated most clearly when examining the differences between roadway types. With small effect sizes, the study was stillable to find highly significant difference at p