Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
9 CHAPTER 3 Evaluation of Offset Left-Turn Lanes: An Investigation Using the SHRP 2 Naturalistic Driving Study Karin M. Bauer and Jessica M. Hutton (MRIGlobal) Introduction The scope of MRIGlobalâs research is to determine whether data from the NDS can be effectively used to determine if offset left-turn lanes affect gap acceptance behavior and improve safety for left-turning vehicles. Results from the full analysis will provide design guidance for offset left-turn lanes that is lacking from the current AASHTO Green Book (AASHTO 2011) and state highway agency design manuals. Left-turn lanes are used at intersections to provide a safe location for storing left-turning vehicles out of the through traffic lanes while their drivers wait for a suitable gap in opposing traffic to turn left. The provision of a left-turn lane minimizes the potential for rear-end collisions with through vehicles approaching from behind the left-turning vehicle and the pressure on left- turning drivers to leave an exposed position and accept an inappropriate gap in opposing through traffic. The Highway Safety Manual (AASHTO 2010) provides crash modification factors (CMFs) for left-turn lanes at four-leg intersections that range from 0.45 to 0.90, depending on the intersection type, area type, and number of left-turn lanes provided. However, left-turn lanes along roadways with medians may create a safety concern, as vehicles in opposing left-turn lanes may block one anotherâs views of oncoming through traffic. A geometric design solution for the sight obstructions that can occur due to opposing left-turn vehicles is to offset the left-turn lanes (i.e., to move the left-turn lane laterally within the median so that the opposing left-turn vehicles no longer block the sight lines of their drivers). The drawings in Figure 3.1 illustrate intersections with negative offset, zero offset, and positive offset for opposing left-turn lanes.
10 Figure 3.1. Illustration of positive, zero, and negative offset left-turn lanes. (Persaud et al. 2009, adapted from Staplin et al. 2001) Offset left-turn lanes are used at signalized and unsignalized intersections. At signalized intersections, they are most effective when permissive-only left-turn phasing is used, since left- turning vehicles do not have to select appropriate gaps for completing their maneuvers during a protected left-turn phase. While the principle of offset left-turn lanes is accepted based on anecdotal evidence, there is no conclusive quantitative evidence of their effects on driver behavior or crash reduction. MRIGlobal has completed a proof-of-concept study using a limited dataset from early in the collection of NDS data. The work included identifying appropriate intersections to study within the NDS areas; developing data requests and assisting in development of queries of the NDS dataset; reducing NDS video data to capture the relevant information for the analysis; and completing a preliminary analysis of the data to demonstrate the methods that would be used for a full analysis with a larger dataset. Working with the NDS Data For this proof-of-concept study, NDS data were requested for left-turn maneuvers at six signalized intersections in the Raleigh/Durham, North Carolina and Buffalo, New York study areas. The dataset provided includes forward-facing and rear-facing videos, basic trip characteristics, and some time-series data for the trip segments that took place within a 1,000-ft radius of the center of the study intersections. In addition, forward-facing videos were received for all NDS study vehicles passing through one of the study intersections within the same 1,000- ft boundaries. The primary source of data for identifying rejected and accepted gaps was the videos from left-turning NDS vehicles.
11 Figure 3.2. Illustration of vehicle maneuvers at an intersection without offset left-turn lanes. Figure 3.2 shows a drawing of an intersection without offset left-turn lanes at which an opposing left-turn vehicle is potentially blocking the left-turning driverâs view of opposing traffic. The vehicles in the drawing are as follows: ⢠Vehicle A is the NDS study vehicle (the instrumentation in Vehicle A collects the data needed for this study). ⢠Vehicle B is making the opposing left-turn maneuver to Vehicle A. ⢠Vehicle C is a through vehicle on the same approach as Vehicle B; Vehicles A and C are potentially in conflict as the driver of Vehicle A considers making a left-turn maneuver; several âVehicle Csâ may pass before the driver of Vehicle A accepts a gap and completes the left-turn. The illustration shows the general field of view for the forward and rear camera installed in Vehicle A. The forward video was used to gather gap rejection information as Vehicle Cs passed through the intersection prior to Vehicle A accepting a gap. The rear video was used after Vehicle A completed the turn to view the next Vehicle C to pass through the intersection, allowing the analyst to record the length of the accepted gap. In addition to gathering this
12 information for Vehicle A, the gap acceptance behavior of vehicles queued in the left-turn lane ahead of Vehicle A can typically be observed in Vehicle Aâs forward-facing camera. These data were also recorded during the data reduction process. Figure 3.3. Screenshot of prototype user interface developed by MRIGlobal in Phase 1. A user interface (Figure 3.3) was designed using LabView software to simplify the data reduction process and ease workload for the analyst. The key characteristics of the interface include: ⢠Automatic population of video file name; ⢠One-click procedure for recording the timestamp of events of interest; ⢠Drop-down menus or buttons to present available event codes to the analyst; ⢠Population of default values where appropriate, so if no change is needed, the analyst can skip the variable field; and ⢠Intuitive design and layout. Preliminary Analysis Table 3.1 presents a summary of the data that were reduced from video files. Distributions of rejected and accepted gaps are shown in Figure 3.4, separately for the three types of offset. A basic logistic model was developed using the data (a) combined across the three offset types and (b) using offset type as a factor to illustrate the methodology proposed in the Phase 1 work plan. The modeling results are illustrated in Figure 3.5 without regard to offset type (left) and using
13 offset type as an analysis factor (right). Each plot shows the functional form that relates the probability of accepting a gap and that gapâs length. From the model represented on the left-hand side plot in Figure 3.5, the research team then estimated the critical gap length, T50, at which the probability of accepting a gap is equal to that of rejecting it. The estimated T50 is 7.8 sec with a 95-percent confidence interval of 7.1 to 8.8 sec. The plot on the right-hand side in Figure 3.5 illustrates the probability curves when using offset type as a factor in the modeling. The analysis of variance associated with this logistic model did not show a significant offset effect. This was expected because the data are based on only six intersections, only one of which has a negative offset. In summary, because of the limited amount of data available in this project, no conclusions could be drawn yet regarding the safety impact of offset left-turn lane design. However, the team met the project goal of showing that the analysis could be accomplished using the NDS dataset. Table 3.1. Data Captured in Phase 1 Unique object ID (VTTI) Offset type Number of left-turn maneuvers recorded Total number of rejected gaps Total number of accepted gaps Number of events with a vehicle present in the opposing left-turn lane NDS vehicles Non-NDS vehicles No Yes 7 Positive 9 2 39 2 20 21 13 1 0 4 0 4 0 19 Zero 32 10 87 10 61 36 23 25 1 34 2 31 4 28 23 14 104 15 81 20 33 Negative 3 0 10 1 11 0 Total 93 27 279 30 208 81 Plans for Phase 2 In Phase 2, the research team plans to increase the sample size by using a dataset of approximately 70 intersections, and to categorize the intersections by offset distance (likely a 3-ft to 4-ft range) as well as offset type. It is anticipated that the 70 intersections will yield between 2,500 and 4,000 trip segments, or left-turn movements completed by NDS study drivers, for analysis. In addition, each trip segment may also include observations of left-turn maneuvers made by non-NDS study vehicles for which gap acceptance can be observed. In turn, each left- turning vehicle will yield up to several rejected gaps and one accepted gap (if it can be observed or estimated from the video data). Based on the small sample in Phase 1 that included just a few months of NDS trip data, it is estimated that the full Phase 2 study, which will include at least three times as many NDS driver trips, may yield in the order of 8,000 rejected gaps and 1,000 accepted gaps. With this dataset, the team will be able to determine if offset left-turn lanes affect driver gap acceptance behavior, as well as whether the presence of a vehicle in the opposing left-turn lane has an impact on the effect. The research will provide design guidance for offset left-turn lanes that is lacking from the current AASHTO Green Book (AASHTO 2011) and state highway agency design manuals. The interpretation of the research results from Phase 2 will establish a
14 minimum desirable offset for opposing left-turn lanes and determine how that information can best be presented as design guidance for application by intersection designers. This should have a direct impact on fatal and injury crashes that involve left-turn maneuvers, as well as on many less severe crashes. Figure 3.4. Distribution of gaps by left-turning study vehicles. Figure 3.5. Probability of accepting a gap as a function of gap duration--offset types combined (left) and by offset type (right).
