U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-HRT-09-061
Date: February 2010
Simulator Evaluation of Low-Cost Safety Improvements on Rural Two-Lane Undivided Roads: Nighttime Delineation for Curves and Traffic Calming for Small Towns
PDF Version (1 MB)
PDF files can be viewed with the Acrobat® Reader®
This experiment investigated visibility enhancements for rural curves and speed–calming treatments for small towns. The curves and towns were combined into a single driving simulator scenario to increase the efficiency of the experiment and reduce boredom for the participants. Consequently, the methodology first describes the elements which were common to both the curves and the towns. It then describes the elements that were specific to either the curves or the towns alone.
The participants received each treatment condition. The curves or towns were separated by a long tangent segment of roadway, and the approach to each curve or town was regarded as an independent trial. It was assumed that there would be little or no carry–over from one trial to the next. Each drive consisted of 26 trials with 20 curves and 6 towns in a quasi–random order separated by a tangent. These tangent segments were 20, 25, 30, 35, or 40 s in duration (driving at 55 mi/h (88.5 km/h)), presented in a uniform random distribution so that each curve or town appeared at a different distance down the road on any given trial. At the beginning of the tangent preceding a town, the simulated driving condition instantly changed from night to day. At the end of the town, the simulated driving condition instantly changed back to night.
Each participant was tested on two different days. The first day consisted of a familiarization drive, a training drive, a practice drive, and a single test drive. The familiarization and training drives were employed so that the participants were acquainted with the driving simulator and became comfortable with handling the car. The practice drive served as a primer for the first test drive, including only the baseline conditions. On the second day, participants completed three more drives. The first drive on the second day was another practice drive, which served as a refresher from day 1. The participants then completed two test drives like the test drive from the first day. The order of particular curve and town treatments was different for each test drive.
The participants were licensed drivers between the ages of 18 and 88 (the mean age was 57.6). They were recruited from the FHWA Human Centered Systems participant database, from word–of–mouth, and from newspaper and online advertising in the greater Washington, DC, metropolitan area. Of the 36 participants who completed the experiment, half were under 65 years of age (the range was 18–64 years old with a mean age of 41.7), and half were above 65 years of age (the range was 66–88 years old with a mean age of 73.6). Each age group (younger and older) was evenly distributed between males and females. The distribution of research participant characteristics is shown in table 1. Although the sample of participants was balanced for age and gender, these factors were not analyzed in the experiment. Participants were given a vision screening test to ensure that they met a minimum visual acuity requirement of at least 20/40 in at least one eye (corrected if necessary). Of the 40 research participants who began the experiment, 4 dropped out as a result of simulator sickness.
The FHWA HDS is a relatively high–fidelity research simulator. Simulator components include a 1998 Saturn SL1 automobile cab and chassis, five projectors, and a cylindrical projector screen. Each projector has a resolution of 2,048 pixels horizontally and 1,536 pixels vertically. The image on the screen wraps 240 degrees around the forward view. Measured horizontally, the projection screen is 9 ft (2.7 m) from the driver's eye point. Under the vehicle chassis, there is a 3 degree–of–freedom motion system which is capable of moving the vehicle approximately ±12 degrees in pitch and roll and ±4 inches (10.2 cm) in heave. A sound system provides engine, wind, tire, and other environmental sounds. Rear view mirrors are simulated using 4.8–inch (12–cm)–high by 7.8–inch (19.7–cm)–wide color liquid crystal displays that have a resolution of 800 pixels horizontally and 480 pixels vertically. A picture of the FHWA HDS is shown in figure 1 .
Figure 1 . Photo. FHWA HDS.
The vehicle dynamics model is calibrated to approximate the characteristics of a small passenger sedan, and data capture is synchronized to the frame rate of the graphics cards (mean rate = 100 frames per second). Data recorded from the vehicle dynamics model includes speed, longitudinal acceleration, lateral acceleration, throttle position, brake force, vehicle position, and heading. A description of the basic simulator system architecture may be found in Advanced Rendering Cluster for Highway Experimental Research.(15) Only speed and longitudinal acceleration were analyzed in this experiment. The HDS has an infrared camera system to monitor the research participants' faces for signs of possible simulator sickness. There is also an intercom system so that the experimenter can maintain verbal communication with the research participants at all times.
Upon arrival, participants read and signed an informed consent form. They were then given a vision screening and a verbally administered health screener. The participants were led to the driving simulator where they were asked to complete a Simulator Sickness Questionnaire (SSQ) and were given a test of postural stability by means of sway magnetometry.(16,17) The same SSQ and postural stability tests were administered after each test drive in the simulator.
Participants were given instructions to read before each driving session. Then, the experimenter reiterated important information and answered questions. Day 1 began with a familiarization drive. This session lasted approximately 3 minutes or until the participants felt comfortable handling the simulator. The participants then took a 5–minute break and completed another SSQ. The next session was curve training where participants drove a series of eight horizontal curves which gradually increased in severity. This session lasted approximately 5 minutes or until the participants felt comfortable negotiating the curves. A 5–minute break and SSQ followed.
Participants progressed to the practice session, which consisted exclusively of the baseline conditions for both the curves and towns. They negotiated eight curves and three towns in this scenario. Participants were instructed to maintain a speed of 55 mi/h (88.5 km/h) on the tangents; however, they could slow to any speed for the curves and towns. They were asked to drive through the curves and towns as they normally would in the real world, to obey the law, and to observe posted speed limits. This practice drive lasted approximately 12 minutes. Upon completion, participants were given a 5–minute break. The first experimental test drive was conducted similarly to the practice drive, and the ordering of conditions in the test drive was randomly assigned before participants arrived. This first test drive lasted approximately 20 minutes.
Participants returned for their second day within a week of their first day. They were asked to complete another baseline SSQ and postural stability test. Participants were given instructions to read before each drive just as on the first day. The experimenter reiterated important information and answered any questions. The first drive, which lasted approximately 12 minutes, was the same as the practice drive from day 1 and served as a refresher for the participants. Upon completion, participants were given a 5–minute break and read the instructions for the next experimental test drive.
Day 2 continued with two experimental test drives, which were the same as the test drive from the first day except that conditions were in a different random order for each drive. The first test drive lasted approximately 20 minutes. Following this test drive, participants were given a short break. They were then given the instructions for the final test drive, including a brief explanation of the meaning of the streaming light patterns. Participants then drove one final test drive in the same manner as the previous test drive. Upon completion of the final test drive, participants were given a final questionnaire, debriefed, and paid for their participation.
All curves and their preceding tangents consisted of two–lane rural roadways driven at night with no fixed roadway lighting. There was no traffic on the roadway in either direction. Although traffic could have been simulated, the glare from oncoming headlights was more difficult to simulate; therefore, it was decided to employ basic driving conditions without any traffic and to depend upon possible future studies to add the complexities of traffic and glare.
Table 2 shows the curve roadway characteristics. The rural tangent and curve roadway segments had lane widths of 11 ft (3.4 m) with 3–ft (0.9–m) paved shoulders on either side. The radius of curvature was either 100 ft (30.5 m) for the sharp curves or 300 ft (91.4 m) for the less sharp curves. The deflection angle was 60 degrees for both types of curves. Both types of curves were quite sharp, and drivers needed to slow down to negotiate either type. Such curves would have posted advisory speeds of 20 or 30 mi/h (32.2 or 48.3 km/h), respectively, but no speed postings were present. The term gentle was used to distinguish the less sharp curves for the research participants. There were an equal number of right–hand and left–hand curves. There was no superelevation on any of the curves. While the simulator could reproduce the effects of superelevation to some degree, the reproduction of these effects was only partial and uncertain; therefore, it was decided not to employ any. The rural scenery on either side of the curves and their approaching tangents consisted of open farmland, stretches of trees, and occasional farm houses or barns. Only a small portion of this scenery was visible to the participants due to the nighttime driving environment. There were no curve warning signs preceding the curves. Advance warning signs were purposely not employed so as to measure the effects of the pavement markings and the PMDs themselves to enhance driver detection of curves ahead. At the beginning of half of the tangent sections following a town, there was a speed limit sign indicating 55 mi/h (88.5 km/h).
Table 2. Curve roadway characteristics.
1 ft = 0.305 m1 mi = 1.61 km
For the curves, the baseline condition consisted of standard 4-inch (101.6-mm)-wide double yellow centerlines on the roadway, both on the preceding tangent and on the curve itself (see figure 2). In the case of the curves, the first low–cost safety improvement beyond the centerlines was the addition of conventional 4–inch (101.6–mm) white edge lines to both sides of the roadway both on the preceding tangent as well as on the curve (see figure 3.)
Figure 2. Screenshot. Curve baseline condition.
Figure 3. Screenshot. Edge lines condition.
All figures (except figure 1) depicting driving scenes represent simulator screen captures that were taken from a higher angle than the driver's eye level in order to better display the features of the treatments. For the nighttime curve scenes, this elevated vantage point also resulted in an unrealistically elevated headlight angle relative to the roadway. Thus, for the curve scenes, the illumination, shading, and reflection patterns differed somewhat from those in the actual simulator scenes viewed by the research participants.
The next levels of improvement involved the application of various configurations of reflectorized PMDs in addition to the 4–inch (101.6–mm) edge lines and centerlines. The first PMD configuration was the standard installation of delineators on the far side of each curve, which is one of the options indicated in the Manual of Uniform Traffic Control Devices (MUTCD).(18) This single side PMD condition is shown in figure 4. The second PMD configuration provided PMDs on both sides of the roadway (not a currently adopted MUTCD option). This PMD condition is shown in figure 5. For both of these configurations, the reflectorized delineators were spaced according to the formula given in MUTCD, and both centerlines and edge lines were present. Thus, up to this point, all curve treatments were additive. From the baseline condition, the first treatment added was edge lines, followed by PMDs on the far side of the curve, and then PMDs on the both sides of the curve.
Figure 4 . Screenshot. Single side PMDs condition.
The third configuration employed similarly spaced PMDs with simulated LED lamps at the top of each post (above the standard reflector panel; see figure 6). These enhanced delineators were located on the far side of each curve. The LED lamps were programmed to create a repetitive streaming light pattern moving in the direction of the road curvature. The LED lights streamed faster for sharp curves and slower for gentle curves. The repetition rate of the streaming light patterns was 3 Hz for sharp curves and 1 Hz for gentle curves. Thus, the LED–enhanced delineators provided information on both the direction and the severity of approaching curves.
In the streaming PMD condition displayed in figure 6, edge lines were also present. The streaming nature of the stimulus could not be conveyed in the static simulator screen capture. In the simulator, the lights were briefly illuminated (for approximately 250 ms) in a sequential manner to create a moving pattern which would traverse the scene from the lower right to the middle left at different rates depending upon the severity of the curve. Such a streaming light technology for rural two–lane curves is not yet mature, but it has some precedent in other countries where it has been applied to curves on limited access roads using only a directional cue.(10) In the current experiment, the streaming lights were operating continuously. They were not activated by the approaching vehicle by means of detecting its headlights, noise, or motion, although such activation might be appropriate for implementation on rural two–lane roads at night.
Figure 6. Screenshot. Streaming PMDs condition.
These five conditions (four treatments plus the baseline) were paired with four combinations of roadway geometry: right curves, left curves, sharp curves (100–ft (30.5–m) radius), and gentle curves (300–ft (91.4–m) radius). This made for a total of 20 unique curve segments in the experiment. All experimental drives contained each of these 20 curves in a different order.
The luminance of the curve treatments, the roadway, and the background scene were measured in the simulator by means of a photometer with a 6–minute spot. Measurements were taken at different simulated scene distances relative to the driver's eye point, but the distance most illuminated by the vehicle headlights in the simulation was taken as the most relevant. This position was 82 ft (25 m) ahead of the driver's eye point. This position would be roughly equivalent to the center of the illuminated circle ahead of the vehicle in figure 6. Table 3 shows the luminance measurements made at that location. At simulated distances greater than 82 ft (25 m), the stimulus luminance was often below the sensitivity range of the photometer.
1 cd/m2 = 0.2919 fl
Participants were instructed to verbally indicate the direction and severity of each approaching curve as soon as they were confident that they could identify the particular roadway feature. They said "right" or "left" when they could predict the direction of the curve ahead followed by "sharp" or "gentle" when they could predict its severity. The only other word that they were allowed to say was "wrong" if they needed to correct the previous response. The task required participants to use whatever information was available to them to predict the direction and severity of the curves as far ahead of the curve as possible.
The output of a separate microphone was recorded by the simulator software system to recognize the onset of verbal responses made by the research participant. The voice onset times for the responses "right," "left," "sharp," and "gentle" were captured, and the distance from this voice onset to the point of curvature (PC) of the curve ahead was computed. This automated system provided feature recognition distances for the curve stimuli. The experimenter recorded the correctness of each verbal response by means of a keypad entry.
Day 2 consisted of a refresher practice drive and two test drives. Following the first test drive, participants were given a short break. They were then given a one–page questionnaire to ascertain how well the participants learned the meaning of the streaming lights on their own. Next, the participants were given the instructions for the final test drive. Included at the end of these instructions was a brief explanation of the streaming light patterns and how these patterns indicated both the direction and the severity of the upcoming curve.
For the town portion of the experiment, a single small town was simulated, and it was approached an equal number of times from each direction. The town was always presented in simulated daylight and consisted of a main two–lane roadway with marked parking spaces on each side. The town was about 1.5 blocks long with an intersection at each end of a straight central block, which was 250 ft (76.2 m) long. Single–story and two–story commercial and residential buildings lined both sides of the central block and extended 50 ft (15.2 m) on either side of the intersections at the entrance and exit of the town. Thus, each town segment was about 450 ft (137 m) long and was preceded and followed by a long rural tangent. There was no traffic on the roadway in either direction. There were sidewalks along each side of the main road within the town limits, painted crosswalks, access ramps at all intersections, and typical traffic signs for a small town. There were no pedestrians in the town. Also, the towns had no speed limit signs either before or in the town, but half of the time, there were 55 mi/h (88.5 km/h) speed limit signs on the town exits to remind the participants to accelerate to 55 mi/h (88.5 km/h) in the long tangent ahead. Table 4 shows the town roadway characteristics. Although speed limit signs were only on half of the town exits, the instructions to the research participants were to maintain a 55–mi/h (88.5–km/h) speed in all long tangent roadway sections.
1 ft = 0.305 m1 mi = 1.61 km
For the towns, the baseline condition consisted of standard 4–inch (101.6–mm)–wide double yellow centerlines on the roadway, both on the preceding tangent and in the town itself. This baseline condition is shown in figure 7 . As was the case for the figures depicting the curve stimuli, an elevated vantage point was employed in figure 7 and in all of the subsequent figures depicting town stimuli. This elevated vantage point is higher than the one employed to portray the curves to reveal the more complicated roadway geometry associated with the town stimuli. The baseline condition had no cars parked in any of the marked parking spaces. By way of contrast, an additional condition was investigated with cars parked in most of the marked parking spaces on both sides of the main road. The parked cars condition is shown in figure 8. The first low–cost safety improvement for the towns, beyond the centerlines, was the addition of bulb–outs at all intersections in the town. These bulb–outs were simulated as curb and gutter modifications. This curb and gutter bulb–out condition is shown in figure 9. In addition, a less expensive bulb–out configuration was implemented by means of pavement markings alone. This painted bulb–out condition is shown in figure 10. These bulb–outs were applied to both main road intersections in the simulated town (four bulb–outs per intersection or eight bulb–outs altogether). These bulb–outs were designed to reduce driver speed through the town without having a significant negative impact on traffic operations. Each travel lane was 11 ft (3.4 m) wide in the bulb–outs.
The next low–cost safety improvement consisted of chicanes at the entrance and exit of the town. These chicanes were first implemented in the standard manner by means of curb and gutter modifications. The chicanes were designed with adequately long taper lengths to achieve appropriate speeds while also allowing for large truck traffic. This curb and gutter chicanes condition is shown in figure 11. The chicanes were also implemented by means of pavement markings only. This painted chicanes condition is shown in figure 12. Each experimental drive contained one each of these six conditions in a different order with three approaches from each direction.
Figure 7. Screenshot. Town baseline condition.
Figure 8. Screenshot. Parked cars condition.
Figure 9. Screenshot. Curb and gutter bulb–outs condition.
Figure 10. Screenshot. Painted bulb–outs condition.
Figure 11. Screenshot. Curb and gutter chicanes condition.
Figure 12. Screenshot. Painted chicanes condition.
Design of Small Towns and Traffic–Calming Treatments
A state–of–practice search was performed to design an appropriate typical roadway cross section for the small town as well as to devise the bulb–out and chicane treatments. The town roadway cross section consisted of two 11–ft (3.3–m)–wide travel lanes and two 8–ft (2.4–m) parking lanes with a 5–ft (1.5–m) sidewalk on each side of the road. The total distance from curb to curb was 38 ft (11.6 m). For determining an appropriate design for the bulb–out, examples were obtained from the Delaware Department of Transportation (DelDOT) as well as from San Diego, CA ; Washington, DC. and Fairbanks , AK . The bulb–out designs from these sources were fairly similar with slight variations. Ultimately, a bulb–out design was developed as shown in figure 9 and figure 10 . The bulb–outs protruded only 8 ft (2.4 m) into the roadway on each side, leaving the full 11–ft (3.3–m) travel lane width in each direction. They extended 16 ft (4.8 m) along the edge of the travel lane and flared back to the parking curb at a 45–degree angle. Thus, the overall extent of the bulb–outs was 24 ft (7.3 m) with curb radii ranging from 2.5 to 6 ft (0.76 to 1.8 m).
Chicane examples were obtained from the DelDOT as well as from Cambridge, MA ; State College, PA ; San Diego, CA ; Washington, DC. and Fairbanks, AK . Unlike the bulb–outs, chicane designs were different from each other in individual cases. Chicanes are often designed for a specific residential or neighborhood situation and not as a general treatment for a main road through a small town; however, State College , PA , implemented chicanes on one of its primary roads through the town. This site provided the basis for the design that was incorporated into the driving simulator (see figure 11 and figure 12). Figure 13 shows a plan view of the curb and gutter chicane. To provide a comfortable yet effective lateral shift when entering the town, the lanes shifted laterally 3 ft (0.9 m) and then 6 ft (1.8 m) in the opposite direction over a total distance of 88 ft (26.8 m). At that point, the lanes remained displaced by 3 ft (0.9 m) before shifting back to match up with their original cross section at the first intersection (see figure 11 through figure 13). The minimum radius of curvature was approximately 325 ft (99.1 m), and the shifts were sufficiently gradual to maintain speeds of approximately 20 mi/h (32.2 km/h). These gradual shifts were also designed to minimize truck off–tracking and to facilitate emergency vehicle use. The painted chicane followed the same geometry. The chicanes were placed at either end of the town before the beginning of the town itself. The beginning of the town without the chicane was located at the beginning of the first parking space shown in figure 12 (traveling from right to left).
Figure 13. Screenshot. Plan view of chicane geometry.
The experiment was conducted in the FHWA HDS. The driving simulator offered a cost–effective method to study the effects of potential safety countermeasures on driver behavior. In addition, field validation is recommended for safe application of countermeasures that have not been previously proven. The current experiment investigated countermeasures for two different safety problems: (1) running off the road on rural curves at night and (2) speeding through small towns. For the nighttime rural curves, two safety surrogate measures were used to infer possible reductions in run–off–the–road crashes: (1) reduced driving speed in the curve and (2) increased curve direction and severity detection distance. The experiment employed rural curves and individual small town segments preceded by a two–lane rural roadway tangent of about 30 s on average. Each participant drove in 3 experimental sessions, each consisting of 20 curves and 6 towns in a different order. Before the final session, the participants were given additional information on the coding cues for the curve condition that employed streaming PMD lights.
The following countermeasures were tested: edge lines, two configurations of standard PMDs, and PMDs enhanced with streaming LED lights. There were four safety treatments and one baseline condition for a total of five types of roadway delineation. The 5 curve conditions were combined with 2 curve directions (right and left) and 2 curve severities (sharp and gentle) for a total of 20 unique curves. Between the second and third test drive in the simulator, the participants were informed how to interpret the enhanced PMD condition with streaming LED lights.
There were five safety treatments and one baseline condition for a total of six configurations of towns. The selected low–cost safety improvements for small towns included bulb–outs, chicanes, and the presence of parked cars along both sides of the road. This latter condition was included to test the hypothesis that parked cars might serve as a speed–calming measure. The bulb–outs and chicanes were simulated in two ways: (1) concrete curbs and gutters and (2) pavement–marking paint.
The measures of effectiveness were vehicle speed and acceleration at various sampling points along the roadway. In addition to speed and acceleration, vehicle position in the travel lane as well as the magnitude and frequency of corrective steering are all important safety surrogate measures when navigating a horizontal curve. It was possible to measure both vehicle position in the travel lane and corrective steering actions in the FHWA HDS. However, these measurements were beyond the scope of this experiment. Time–stamped voice responses were also recorded to indicate driver feature detection for both the direction and severity of curves ahead in the roadway. In addition, questionnaire responses were obtained after certain driving sessions.