Using Driving Simulator for Advance Placement of Guide Sign Design for Exits along Highways

Using Driving Simulator for Advance Placement of Guide Sign Design for Exits along Highways Fengxiang Qiao, Xiaoyue Liu, and Lei Yu Department of Transportation Studies Texas Southern University 3100 Cleburne Street, Houston, TX 77004, USA qiao_fg@tsu.edu Abstract One of the most important and effective ways to smoothly guide drivers, especially the outlanders, is the advance placement of guide signs. The current MUTCD provides general guidelines in deploying guide signs but does not give a methodology to determine the advance placement distance of the guide signs taking important variables into consideration. This research proposes an operational procedure using driving simulator to estimate the advance placement distance for guide signing. A probability-based model to describe the traffic flow around the subject vehicle has been employed, while its parameters are calibrated based on the field data. The ambient vehicles are coded accordingly in the driving simulator through different simulation scenarios. Test participants are exposed to different traffic conditions and advance placements of signs. The experimental design focuses on investigating different lane positions of the subjects, lanechanging behaviors, and traffic flow rates that affect the placement of the sign. The participants are asked to complete the test questionnaires, and the results are recorded in two categories: (1) subject responses to the effectiveness of the advance placement, and (2) objective records of driver s choices at exits. These data are further analyzed with various combinations of independent and dependent variables designed. Based on the analytical results, the optimal advance placement of signs is recommended that considers the drivers various behaviors and their physical location on roadway. Résumé Dr. Fengxiang Qiao is an Assistant Professor of Department of Transportation Studies, the director of Center for Modeling and Simulation (CMS), and a leading researcher of Center for Driver Behavior (CDB), Texas Southern University (TSU). Dr. Qiao s research interests include driving behavior study, traffic and transportation simulation, intelligent transportation forecasting and simulation, intelligent data processing and archiving. In recent years, he has published more than 60 academic journal papers, peer-reviewed conference papers, and research reports.

Introduction Drivers, especially those who are not familiar with the specific highways that they use, would rely on guide signs for information such as interchange approaches, names of streets or freeways, upcoming major traffic generators, etc. Advance guide signs, which notify drivers of the interchanges or ramps, play important roles on freeways. The lack of guide signs may cause problems if drivers do not recognize a necessary lane change early enough (Upchurch, 2005). The 2003 Manual on Uniform Traffic Control Devices (MUTCD, 2003) offers standards, guidance, and options for freeway guide signing, but it did not provide systematical methodology to determine placement of signs. For example, for major and intermediate interchanges, it indicates that advance guide signs should be placed 1 km (or 0.5 miles), and 2 km (or 1 mile) in advance of an exit. However, it does not offer guidelines for installing guide signs within 0.5 miles from exits. Recently, some pilot studies of sign and signal placement have been initiated. Zwahlen (2003) evaluated the placement of diagrammatic on-ramp advance guide signs based on a field test. Li (2006) proposed a geometric-relation-based operational model for advance sign placement. Wang (2007) presented an analytical model relating the ramp signal placement with geometry, speed, and drivers characteristics. However, these approaches did not consider the dynamic driving behaviors, and thus are not capable of effectively placing advance guide signs. The objective of this research is to use driving simulator to determine the placement of advance guide signs and arrow exit sign, and to identify the sight distances for road signs and factors that may limit drivers reaction time. This information will further decide the effectiveness of the placed signs. Driving Simulator The Driving Simulator on a fixed base at Texas Southern University (TSU) is a fully integrated, immersive driving simulation system of high fidelity. It is designed for ground vehicle research and training applications. It provides multi-channel audio/visual systems, with 180, 240, 300 and 360 wraparound display options, a full-width automobile cab including windshield, driver and passenger seats, central console, dash and instrumentation, plus real-time motion simulation through Q-Motion platform. Drivers have control of the acceleration pedal, brake pedal, and the steering wheel, exactly like what they do in a real automobile. The scene transmitted onto the screen board is refreshed 60 times per second. Data about driver s position, velocity and brake pressure can be collected easily. This system allows researchers the ability to collect meaningful data about driver performance in a variety of different driving situations (HyperDrive, 2007). The whole experiment was conducted in the Driving Simulator at TSU (See Figure 1). All of the geometric characteristics of the freeway and guide signings are re-created using HyperDrive TM software to demonstrate a simulated environment. This allows the visual

driving experience through the simulated environment to be a replication of authentic driving experience, as on a real site. Figure 1 The TSU owned driving simulator DriveSafety DS-600c Methodology Description The proposed methodology evaluates the placement of advance guide sign for exits along highways, taking into consideration critical variables including traffic flow rate, number of lanes, designed speeds of main lanes and exit lane, number of exits within one mile, etc. (See Figure 2) Traffic Flow Heavy Traffic Light Traffic No. of Lanes 3 4 5 3 4 5 No. of Exits within one mile 0 1 0 1 0 1 0 1 Designed Speed 55mph 60mph 65mph Figure 2 Variables affecting the placement of advance guide sign For different combinations of highway geometry and characteristics, the scenarios in driving simulator are designed and tested by subjects. Data related to subject driver s trajectory can be collected in the driving simulator. The data needed can be classified as objective one and subjective one (Table 1). Table 1 Data Collected Data Type Items Correct Response Objective Data Distance during lane change from leftmost to rightmost lane Distance between the sign and driver clearly sees the sign (programmable button) Subjective Data Survey Result The research would distribute different weights to each item based on different concerns. With all the information, a conclusion can be drawn to determine the optimal alternative for the placement of sign. Case Study As illustrated, the case study demonstrates a procedure of using driving simulator to determine the placement of advance guide sign. The case study is a simplified version of the methodology, which only includes one variable (traffic flow).

Two experiments were designed. Experiment 1 was to determine the necessity to place a guide sign 1/4 mile ahead of the exits. Experiment 2 was to determine the appropriate distance from exit where the arrow exit sign should be installed. With different experiments, this research analyzed the effect of heavy and light traffic flow. Based on the Level-of-Service (LOS) criteria 1 for multilane highways (Highway Capacity Manual, 2000), the free flow speed for the experiment is 60mph. The heavy traffic flow was selected as level D, and a light traffic flow was defined as level B. Tests were conducted under both flow statuses so as to incorporate the influence of traffic flow variations. Micro-simulation Process Traffic Flow Generation To achieve the utmost sense of reality, the simulating system first generates a series of backdrop vehicles around the subject vehicle with certain speed. The distribution of the headways between the generated vehicles follows negative exponential distribution. Specifically speaking, the vehicles will be generated at an interval h, calculated by the following equation: h = ( H avg hmin )[ ln(1 R)] + H avg hmin (1) where, h is the time interval; H avg is the average headway; h min is the minimum headway; and R is a random number in the range of 0.0 to 1.0. The different vehicle types are also taken into consideration. The percentage of heavy vehicles is counted from U.S. Highway 101 (Hollywood Freeway) in Los Angeles, California collected between 7:50am and 8:05am on June 15, 2005 (NGSIM, 2005). Vehicle Movement The vehicle movements are controlled by a dynamic model, which consists of two parts. One is the determination of the vehicles accelerations for each frame; the other is the lane change model. The acceleration is determined by the vehicle dynamics based on the desired acceleration and other factors such as current velocity and distance to lead vehicle. Scenario Design Experiment 1 A 3-lane freeway with an exit is simulated in Experiment 1 containing two scenarios. Scenario 1 is around N. MacGregor Way exit, with advance guide signs of 1 mile, half mile and a quarter mile; Scenario 2 is around the Kirby Dr. exit, with advance guide signs of 1 mile and half mile (See Figure 3 (a) and (b) ). Subjects would drive under these two scenarios with a traffic flow condition of LOS B. Experiment 2 1 Level-of-Service (LOS) criteria is defined as six levels from level A to F with A as the free flow status, and F the jam status.

The simulated freeway segment is a 3-lane T- interchange in Houston, with a fourth lane appearing 520m before the exit. Three different horizontal positions of the Almeda Rd. exit signs (near the exit point, 1/4 mile from the exit, and 3/8 mile from the exit) were tested (See Figure 3 (c)). For each placement, two traffic flow conditions (light, heavy) were designed for a comprehensive comparison. Besides the arrow exit sign, 1 mile and half-mile advance guide signs were installed as well, based on the real situation. Alternative 1 Alternative 2 Alternative 3 (a) Scenario 1 in Experiment 1 (b) Scenario 2 in Experiment 1 (c) Alternatives in Experiment 2 Figure 3 Scenarios designed in Experiment 1 and 2 It was hypothesized that the visibility of each sign had something to do with the lane in which the driver was located, and the cognitive demands on the driver. In order to determine the effect of these factors on driver performance, the following items were indicated: 1. The driver s starting lane - The initial position of driver was the leftmost lane, while the tested signs were all indicating the exit on the right. 2. Cognitive demands - The subjects were told to press the programming button when they clearly recognize the guide signs. Subject Test Experiment 1 Twenty-four test subjects participated in Experiment 1 (aging from 23 to 45 with 18 males and 6 females), and were divided into 2 groups. The scenario with a quarter-mile sign was tested by group one; the one without such sign was tested by group two. Participants were instructed before the experiment to exit at N. Macgregor Way (with the quarter-mile sign) and Kirby Dr. (without the quarter-mile sign). All participants were not familiar with the exit sign sequence or geometry before the tests. The operational variables and coordinates of the test vehicles when subjects pressed the button were taken down, and the number of missed or late exits was recorded. Subjects were asked to fill out a posterior questionnaire immediately after the completion of the driving. Experiment 2

Twenty-four test subjects participated in Experiment 2 (aging from 23 to 45 with 18 males and 6 females). For each of the three placement alternatives, there were heavy and light traffic flow situations. Test Result and Analysis Experiment 1 General Error For the freeway test with the quarter-mile sign, one person did not have enough reaction time to take the correct exit. For the one without quarter-mile sign, 3 subjects failed to take the correct exit. While due to the limited sample size, one or three failure(s) of taking a designated exit may not directly imply whether a guide sign is necessary or not, the analyses of the performance of these failure vehicles details could be helpful to even large sample tests in the future. Lane Position When Subjects Pressed the Button Table 2 shows the positions of subjects when they pressed the button. For the test with quarter-mile sign, when subjects pressed the button to report their recognition of 1/4 mile sign, 8 of the total 12 subjects were already in the rightmost lane. For the one without quarter-mile sign, 7 of the total 12 subjects were in the rightmost lane when they saw the half-mile sign. This means that for drivers who are not familiar with the specific highway, they either tend to drive in the rightmost lane for any unexpected exit(s) no matter whether a quarter-mile sign is there or not; or they have obtained enough information from the 1 mile, 1/2 mile advance guide signs. Table 2 Number of subject vehicle in different lane positions lane position when see 1/2 mile sign lane position when see 1/4 mile sign Scenario leftmost middle rightmost leftmost middle rightmost Scenario 1 1 5 6 0 4 8 Scenario 2 0 5 7 N/A Table 3 shows the lane position of the subjects who failed to take the correct exit. The error data for Scenario 2 indicates that, even after the subject (ID R in Table 3) saw the 1/2 mile sign, he/she did not change to the exit lane (lane 3). This means either the advance guide signs did not have much effect on the subject, or the subject thought there would be enough time to change lane later and exit successfully. Table 3 Lane position of subjects who failed to take the exit Error Data lane position when see lane position when see lane position at exit point Subject ID 1/2 mile sign 1/4 mile sign Scenario 1 2 2 3 B Scenario 2 3 3 C Scenario 2 2 N/A 2 R Scenario 2 2 3 E Speed Influence The average speed of subject vehicles conducting the test was 59.23 mph (standard deviation 3.95mph), under a speed limit of 60 mph. And there were 4.5% of heavy vehicles among the traffic flow. Table 4 shows the mean speeds and the speeds when erred subjects pressed button. Both subject C and R did not change their lanes when driving between ½ mile and the

exit. Using a Student's-distribution with degree of freedom 11 for all speed data, the 95% confidence interval is [57.03, 61.85], while the 99% confidence interval is [54.79, 64.51]. So the mean speeds of subject C and R not only exceed the speed limit 60 mph, but also are out of the 99% and 95% confidence intervals, respectively. Therefore, speed, instead of the ¼ mile guide sign, most likely plays an important role in their failures toward the correct exits. By eliminating the influence of speed, it is possible that there is no noticeable difference between Scenario 1 (with ¼ mile guide sign) and Scenario 2 (without ¼ mile guide sign). Further larger scale tests on whether ¼ mile guide sign is truly necessary are recommended. Table 4 Speed of subjects who failed to take the exit Error Data Lane positions shift Subject Mean SpeedSpd. when press button for ½ mile guide sign from 1/2 mile to exit point ID Scenario 1 2->3 58.78mph 57.90mph B Scenario 2 3->3 68.25mph 66.79mph C Scenario 2 2->2 63.84mph 62.89mph R Scenario 2 2->3 58.52mph 58.70mph E Survey Result Analysis The posterior survey aims at collecting subjects feeling after experiments, which were then combined with test results for a comprehensive conclusion. The survey shows that 15.4% of the subjects thought the quarter-mile sign was helpful, 30.8% of them thought it was somewhat helpful, while 46.2% subjects thought this sign would be more helpful under much complicated situations, such as there are more exits within one mile, or more lanes. 7.6% of the respondents thought it was not helpful at all. For the specific situation of 3-lane freeway with one exit, quarter-mile sign might not be essential. However, this needs the validation from large scale sample tests in the future. Experiment 2 General Error For the arrow exit sign near the exit point in light traffic, 1 person failed to take the correct exit; and for the one in heavy traffic, 2 people failed to exit. For the arrow exit sign installed in the 1/4 mile scenarios, all the subjects took the correct exit. For the arrow sign installed at 3/8 mile, 1 person failed to take the correct exit under light, and 1 failed under heavy traffic. Similar to Experiment 1, the analyses of these limited failure vehicles cannot definitely conclude the necessity of arrow signs, but, they could be helpful to larger sample tests in the future. Lane Change Conducted After Subjects Saw the Arrow Exit Sign Figure 4 shows the lane change conducted after the subjects recognized the arrow exit sign. The assumption here is that, the more people made lane change after they saw the sign, the more effective the sign would be. However, the number of lane change is not the sole factor that affects the effectiveness of placement. Since all the subjects were unfamiliar to the freeway segment, after the fourth lane appeared (520m from the exit, see Figure 3(c)), the lane change that flowed vehicles into lane 4 (the new rightmost lane,) could be an indicator that this alternative of placement is more effective.

Lane Change Conducted After Saw the Arrow Sign No. of Subjects 4.5 3.5 4 2.5 3 1.5 2 0.5 1 0 Alternative Figure 4 Lane change conducted after subjects saw the arrow sign (A= Alternative) Table 5 lists the destination lanes under light traffic. In Table 5 although alternative 3 showed the most number (4) of lane changes, only 2 of them were directed to the fourth lane. Thus, if the T-interchange is a 3-1 split type, these two people would further miss the exit. Also, alternative 3 is the placement of arrow exit sign around 3/8 mile from the exit. The subjects saw it right after the 1/2 mile sign, so it may be under the illusion that there is still a long way to the exit. Table 5 Destination lane of lane change under light traffic Alternative Light Traffic No. of Lane Change Change to the fourth lane Alternative 1 2 1 Alternative 2 3 3 Alternative 3 4 2 Influence of Speed Table 6 shows the speed information under each scenario. It indicates that within the same traffic situation (light or heavy), the speed of the subject vehicles did not show much differences among the alternatives. Table 6 Speed information for each scenario Traffic Conditions Alternative Mean Speed Standard Deviation Alternative 1 57.20mph 6.35mph Light Alternative 2 58.99mph 6.49mph Alternative 3 57.94mph 5.44mph Alternative 1 44.11mph 7.24mph Heavy Alternative 2 45.26mph 4.96mph Alternative 3 45.88mph 8.12mph Survey Result Analysis The posterior survey indicates that all the subjects thought the arrow exit sign was helpful in providing the exit information, except for the difference in extent (31% thought it was very helpful, 46.2% thought it was helpful, and 22.8% thought it was somewhat helpful). The placement preference of the arrow exit sign was in accordance with experimental result (23.1% thought it would be installed just at the exit point, 53.8% preferred the instalment around 1/4 mile from the exit, another 23.1% thought it would be installed around 3/8 mile from the exit).

Considering the survey with test results jointly, for the specific situation of 3-lane freeway T-intersection type, it is potential that arrow exit sign can be installed around 1/4 mile from the exit. Further conclusions would be based on future large sample tests. Conclusion 2 This paper presents an experimental procedure to evaluate the placement of guide signs. While the test results may not be comprehensive due to the limited sample size, it provides a repetitive procedure to determine the sign placement. This research involves two experiments: one for determining the necessity of installing the quarter-mile advance guide sign on highway; the other one for determining the placement of arrow exit sign along highway. A comparison among different scenarios under various traffic conditions and sign placements was conducted. It is concluded that for the 3-lane highway with one exit within one mile, there may not be too much necessity to install the 1/4 mile guide sign. And the arrow sign can potentially be installed at around 1/4 mile from the exit point. In this research, twenty-four test subjects participated in both experiments, with one to three subjects missed the correct exits. On the whole, this preliminary study with small sample size tests will lead to the knowledge for designing large sample experiments to further recommend optimal sign placements based on driver s comprehension. Reference 1. HyperDrive & Vection User's Guide Version 1.9.35, DriveSafety Inc., 2005. 2. FHWA. http://www.ngsim.fhwa.dot.gov/ Last Accessed June 08, 2007. 3. Li, J., C. Lan, D. Chimba, A Supplement to Advance Guide Sign Placement Guidelines in MUTCD. In Transportation Research Board 85 th Annual Meeting compendium of papers, CD- ROM. Transportation Research Board of the National Academies, Washington, D.C., 2006. 4. Li, J., C. Lan, D. Chimba, A Supplement to Advance Guide Sign Placement Guidelines in MUTCD. In Transportation Research Board 85 th Annual Meeting compendium of papers, CD- ROM. Transportation Research Board of the National Academies, Washington, D.C., 2006. 5. Manual on Uniform Traffic Control Devices for Streets and Highways, adopted by the Federal Highway Administration, 2003. 6. Upchurch, J., D. L. Fisher, and B. Waraich. Guide Signing for Two-Lane Exits with an Option Lane: A Human Factors Evaluation. In Transportation Research Board 84 th Annual Meeting compendium of papers, CD-ROM. Transportation Research Board of the National Academies, Washington, D.C., 2005. 7. Wang, Z., Placement Design of Ramp Control Signals, In Transportation Research Board 86 nd Annual Meeting compendium of papers, CD-ROM. Transportation Research Board of the National Academies, Washington, D.C., 2007. 8. Zwahlen, H. T., A. Russ, J. M. Roth, T. Schnell, Evaluation of the Effectiveness of Ground Mounted Diagrammatic Advance Guide Signs for Freeway Entrance Ramps. In Transportation Research Board 82 nd Annual Meeting compendium of papers, CD-ROM. Transportation Research Board of the National Academies, Washington, D.C., 2003. 2 The authors acknowledge the sponsorship of TxDOT research project 0-5800, the project director Ismael Soto, test participants, the language editor Dr. Haiqing Sun, and other personnel that contribute to this research directly or indirectly.