Impact of Driver Information System on Traffic ...

3 downloads 0 Views 865KB Size Report
Nov 10, 2005 - Bottleneck Active 380-390. 0 = Not in Service. 1 = Attention. 4 = Congestion. 5 = No Passing “Trucks Only”. 2 = Slippery Road. 3 = Road Work.
12th World Congress on ITS, 6-10 November 2005, San Francisco

Paper 1003

Impact of Driver Information System on Traffic Dynamics on a German Autobahn: Lessons for U.S. Applications Robert L. Bertini,1* Klaus Bogenberger,2 and Steven Boice1 1. Department of Civil & Environmental Engineering Portland State University P.O. Box 751 Portland, OR 97207-0751, USA Phone: 503-725-4249 Fax: 503-725-5950 Email: [email protected] 2. BMW Group, 80788 München, Germany Phone: +49-(0)89-38230764 Fax: +49-(0)89-38241136 Email: [email protected]

ABSTRACT This paper presents results of an empirical study focusing on the fusion of traffic data from multiple sources—including fixed freeway sensors, floating probe vehicles, dynamic navigation system, and variable speed limit/dynamic message sign data—from a freeway corridor in Munich, Germany. The analysis includes identification of recurrent freeway bottlenecks using archived sensor data, has used these diagnoses to measure the capacity at these cross-sections, and has identified driver behavior (including truck passing activity) that led to the formation of the upstream queues. These results were validated using high-resolution spatio-temporal location data from probe vehicles. The analysis has also compared the magnitude and extent of congestion identified from the archived sensor data with the dynamically reported extent of congestion from the traffic management center. Finally, the study has analyzed the variable speed limit and dynamic message sign data in order to develop an improved understanding of the impacts of these messages on driver behavior and bottleneck formation/location. The results of this project should contribute toward improved travel time estimation and forecasting for traffic management, traveler information and for driver assistance systems. The project benefits from increasing data availability and from the development of a reliable methodology. KEYWORDS: Traveler Information, Variable Speed Limit, Traffic Flow INTRODUCTION This paper presents results from a study that has sought to diagnose bottleneck activations along a 16-km section of northbound Autobahn 9 (A9) in Munich, Germany (see Figure 1). The analysis tools used in this study were curves of cumulative vehicle arrival number versus time and curves of cumulative time-averaged velocity versus time. [1, 2, 3] The archived inductive

Figure 1 – Northbound Autobahn 9 Site Map

loop detector data required to construct these curves were available at one-minute levels of aggregation by lane and vehicle type (autos versus trucks). A dual loop configuration was used to directly measure vehicle speed. This method made it possible to identify several bottleneck activations located near a busy on-ramp during the afternoon peak period on multiple days. [4] This bottleneck’s location and discharge flows were reproducible from day to day. Bottleneck discharge flows also appeared to be relatively stable. BOTTLENECK ACTIVATION Figure 2 shows a speed contour map of the northbound A9 motorway for Thursday June 27, 2002. As shown, a bottleneck was activated twice between detectors 380-390 and once between stations 390 and 420. Detailed analysis confirmed the activation and deactivation times as well as tracking the queue propagation upstream. [4] The analysis initially focused on the bottlenecks in the region between stations 380 and 420 that were present for longer time periods. The disturbance that was visible between stations 320 and 340 from 17:35-17:40 is very interesting. Our analysis has shown that there was a surge in flow at the Garching-North off-ramp at approximately 17:29. The (380-390) average bottleneck outflow was estimated to be 5370 veh/hr during 11 activations over six days. When the bottleneck activations were preceded by freely flowing traffic, the mean sustained flow was 5620 veh/hr, representing an average 4.7% reduction in flow. This is consistent with past research that has found flow reductions upon bottleneck activation. [e.g., 3] Because trucks have different performance characteristics than autos, a detailed analysis of truck

0 20

40 60

80 100 120 140 160

17:28 17:35 15:21

19:40

18:44 17:38

15:24

17:40

15:34 15:41 15:42 15:47 15:58

15:00

16:00

17:00

18:00

19:00

20:00

Figure 2 – Northbound Autobahn 9 Speed Diagram June 27, 2002

2

BN

flow and velocity patterns was conducted just upstream of the bottleneck, with a focus on the 15:21 activation. Surges in truck counts were observed in both the center lane and shoulder lane at approximately the same time the queue passed station 390. The truck speeds dropped in each lane beginning at 15:17, when the average truck speed dropped from 100 km/hr to 80 km/hr in the shoulder lane, followed one minute later by the truck speed drop in the center lane, and in the median lane one minute later. Notable truck movements arose just before and upon the bottleneck’s activation. Truck counts and speeds were also examined at station 420, 1.2 km upstream of the bottleneck. This revealed that trucks remained to the right, on both the München-North on-ramp and on the mainline. Several minutes before the queue formation, there were three surges in truck counts. First, over a two minute period, 9 trucks passed station 420 in the mainline median lane. Next a 120% truck flow increase was observed in the mainline shoulder lane, and almost simultaneously a 190% increase in the truck counts was observed in the right lane of the on-ramp. VARIABLE SPEED LIMIT SYSTEM DATA ANALYSIS This section of Autobahn 9 not only has an extensive surveillance system, but also contains a variable speed limit (VSL) and congestion level information system that uses variable message signs (VMS). Figure 1 shows that there are five variable message sign structures along northbound A9 that are used to display variable speed limit (VSL) information as well as driver information. Analysis of variable speed limits focused on using a time-space coordinate system and comparison to actual measured speeds over the same freeway segments, before, during and after bottleneck activation. The VSL system uses real time data provided via the inductive loops, closed circuit television (CCTV) and remote weather information systems (RWIS) to evaluate the existing traffic and/or weather conditions in order to determine a recommended speed for a particular autobahn segment. The recommended speed is indicated to vehicle drivers using the VMS located along the roadway on overhead gantries as shown in Figure 3. Currently, VSL is implemented on this section of Autobahn 9. In Europe, VSL is applied in order to reduce crashes, reduce congestion, and increase flow during peak periods. Research is being conducted to determine whether deployment on U.S. freeways should be considered. In order to do this, an understanding of when speed limits should

Figure 3 – Variable Speed Limit Gantry

3

be reduced/increased and by how much must be known. The effects (increased flow, safety, and driver compliance) of the speed change on the road network must also be known to determine system benefits. Variable speed limit data for the 5 VMS gantries in the northbound direction were analyzed. Each gantry can display a speed limit “Zulässige Höchstgeschwindigkeit” for each lane and/or a warning or prohibition accompanied with textual information. German laws require that the same speed be displayed for all lanes. Data consisted of the speed limit posted on each sign throughout the day as well as any warnings and/or prohibitions posted on each sign. Speeds posted on the sign ranged from 60–120 km/hr, at 20 km/hr steps. The warning/text signs usually displayed symbols with the following meanings: • • • • • •

“Stau” – Traffic Jam “Staugefahr” – Danger of or high probability of traffic jam “Überholverbot LKW” – No passing for trucks “Nässe” – Slippery road “Ende LKW Überholverbot” – End of no passing for trucks “Ende Überholverbot” – End of all prohibitions

380-390

Bottleneck Active 380-390

120

100

80

BN Legend

120 100 80 60

15:00

16:00

17:00

18:00

19:00

20:00

Figure 4 – June 27, 2002 Northbound Variable Speed Limit Figure 2 shows the speed contour for June 27, 2002 northbound A9, with bottleneck activation and deactivation times annotated. A plot of the variable speed limit as shown in Figure 4 for June 27, 2002 northbound illustrates the varying speed limit designation by the use of shading the

4

designated speed in time and space. Dark shades represent low speeds (60 km/hr) and lighter shades represent higher speeds (120 km/hr). The white region indicates where no speed limit was posted (or if a communication failure occurred during data extraction). The first dashed vertical line represents the activation of the bottleneck between detectors 380-390 at 15:21. The second dashed vertical line indicates the de-activation of this bottleneck at 17:38. The final dashed vertical lines represent the re-activation of this bottleneck between 18:44-19:40. As one can notice by comparing Figure 4 to Figure 2, the variable speed appears to be related to actual traffic conditions or vice versa as seen by the trend in reduction of speeds after bottleneck activation and increase in speeds upon bottleneck deactivation. Message sign AQ 305, upstream from bottleneck location, fluctuated between speeds of 120 km/hr and 100 km/hr beginning in the morning peak until a posting of 80 km/hr was posted at 14:44, 37-minutes prior to bottleneck activation. A speed of 60 km/hr was then posted at 15:26 followed by no speed posted at 15:35. This time marks when the congestion or “Stau” indication was presented to drivers and is indicated by the white region in Figure 4. Comparing changes in posted speeds for message sign AQ 305 with actual traffic conditions for detector station 420 indicated that speeds were generally reduced upon previous measurements consisting of decreases in flow and speed, and increased with previous measurements of increases in flow and speed. It was determined that the speed is adaptive to actual traffic conditions and may have led to the dampening of congested conditions. To further explore the VSL system, vehicle compliance was measured. In order to determine actual vehicle speeds as compared with that of the variable speed limit, the variable speed limit indicated on each gantry was subtracted from the nearest upstream/downstream average loop detector speeds for all vehicles over all lanes. The difference in speed in time and space for June 27, 2002 are shown in Figure 5. The same color reference system as mentioned in the speed Bottleneck Active 380-390

100

BN

80 60 40 20 0 -20 -40 -60 -80

14:30

15:00

15:30

16:00

16:30

Figure 5 – Speed Contour (Actual Speed – VSL)

5

contour plot is used. Prior to bottleneck activation at 15:21, average vehicle speeds over all travel lanes at detector station 390 revealed that speeds were on average 1 km/hr above that indicated on message sign AQ 305 located upstream of this location. This indicates a high percentage of vehicle compliance and that average vehicle speeds were consistent with the speed Bottleneck Active 380-390

380-390

40

BN

20 0 -20 -40 -60

15:00

16:00

17:00

18:00

19:00

20:00

Figure 6 – Speed Contour (Actual Truck – 80 km/hr) indicated overhead. Average trucks speeds were found to be on average 6 km/hr below the posted speed for the same time period. More analysis is underway to determine whether the indicated speed adapted to actual vehicle speeds or if speeds adapted to the speed indicated. In Germany, the regulatory speed limit for trucks is 80 km/hr at all times. With the availability of segregated truck data, average truck speeds over detector stations were able to be compared to determine compliance. Figure 6 reveals that the majority of trucks under free flow conditions traveled +10-20 km/hr above the 80 km/hr regulatory speed on June 27, 2002. In fact prior to bottleneck activation average truck speeds over all travel lanes was found to be 12 km/hr greater than the regulatory 80 km/hr limit. This plot again uses the similar shading method as the speed contour plot and contains data from subtracting the 80 km/hr speed limit from actual truck speeds over each detector station throughout the corridor. Trucks were therefore traveling at speeds greater than their regulatory speed yet less than the speed indicated on the variable message signs. This could cause disruption in the mainline due to trucks maneuvering into the median lane at slower speeds than surrounding traffic, thus creating a wall of slower moving traffic which in turn could prevent vehicles wanting to travel faster from doing so. Next, warnings and prohibitions indicated overhead on the message signs were plotted in time and space to see how they interacted with actual traffic conditions. These warnings and prohibitions are displayed on signs located between speed signs. Symbolic characters are typically used to notate restrictions in replace of textual indication on these signs. Textual 6

information for drivers is located on an additional sign located just underneath these signs. The warnings and prohibitions for June 27, 2002 northbound are shown in Figure 7 and textual information presented to drivers that accompanied the warnings is displayed in time and space in Figure 8. The warnings are labeled in Figure 7 accompanied with symbolic characters used to describe these warnings. The dark shaded region in Figure 8 illustrates the text “Stau” and the lighter region “Staugefahr.” As shown, the no passing for trucks prohibition was displayed throughout the corridor beginning prior to 15:00. This is likely to prevent trucks from changing lanes in high flow conditions because they are required to travel at slower speeds than the remaining traffic stream. As previously mentioned slower moving vehicles into the median lane could create large disturbances in the mainline flow under saturated conditions. Concentrating on message sign AQ 305 this restriction remains active until “stau” or traffic jam is displayed at 15:35. This is 14-minutes after bottleneck activation between detector stations 380-390. It is this time that the system has recognized queued conditions with representation of high flows and low speeds. Looking back to Figure 4, at 15:35 the variable speed limit is no longer indicated on upstream message signs as illustrated by the white shaded region. This is when the system recognized the state of jammed conditions and indicated to drivers that this state was prevalent and that vehicles speeds were already much lower than that displayed. FLOATING CAR DATA ANALYSIS Going beyond data from fixed sensors, data from vehicles equipped with a global positioning Bottleneck Active 380-390

380-390

1 Congestion 2 Attention to Congestion

380-390

Bottleneck Active 380-390

BN Speed Contour Colorbar (km/hr)

7 6 5 4 3 2 1 0

BN

120

100

15:00

16:00

80

60

18:00

17:00

19:00

20:00

Figure 8 – June 27, 2002 Northbound Textual Information

40

BN

20

Floating Car

16:45

16:50

16:55

17:00

17:05

17:10

17:15

17:20

17:25

17:30

Colorbar (km/hr)

15:00

18:00 Analysis 19:00 17:00 Car Data 16:009 – Floating Figure June 27,20:00 2002

6 = End No Passing for Trucks 0 = Not in Service 2 = Slippery Road 4 = Congestion 7Only” 7 = End All Prohibitions 5 = No Passing “Trucks 3 = Road Work 1 = Attention

Figure 7 – June 27, 2002 Northbound Warnings/Prohibitions

system (GPS) were also analyzed. Data consisted of distance traveled (km) and speed recorded every 1-second. One travel run in the northbound direction along Autobahn 9 between detectors 630-260 from 16:50-17:10 experienced congested conditions on June 27, 2002. The trajectory of the vehicle traveling in the evening peak period passed through congested conditions and is shown in Figure 9 (the color is opposite to the speed contour diagram, with dark for high and light for low speeds). As shown, the congested conditions experienced by the floating car added 0:11:33 of travel time as compared to an additional floating car which traveled through the same section experiencing no congestion between 11:55-12:03. Further analysis was conducted to determine the speed and travel time between detectors to determine where the highest levels of congestion were present. The floating car reached station 630 at 16:50 when congested conditions were prevalent. This is noticeable in Figure 9. The floating car experienced the highest level of congestion between detectors 580-560 where it averaged a speed of 12 km/hr over a distance of 0.41 km. Between these two detectors is a busy on ramp which may influence the vehicle delay. The floating car traveled between these detectors between 16:56–16:58. The measured loop detector speeds during the same period were: 35 km/hr (detector 580) and 34 km/hr (detector 560). Congestion could also be highest at this location due the lane drop in the freeway just downstream. ASSIST DATA ANALYSIS Traffic information was provided to BMW drivers via the ASSIST system, an in-vehicle component of the navigation system. Congestion parameters were provided in the form of kilometer-kilometer marker over a duration of time. Two types of messages (types 4 and 5) were deployed for this particular day in the northbound direction. Type 4 messages were generated based on raw inductive loop detector data while type 5 messages incorporated the use of a traffic flow model. For this, traffic data was input into a modeling package with the outputs of simulation congestion parameters provided to the driver. With the availability of an actual and modeled output, both types of messages were analyzed to determine how they compared with respect to one another and to actual traffic conditions as provided by inductive loop detector data. Since messages were provided in both time and space, the estimated congestion parameters provided by the ASSIST system were plotted as rectangles over the speed contour plot. Messages of type 5 for June 27, 2002 northbound are shown in Figure 10. This allows one to visually see the comparison of messages as compared to actual traffic conditions. As can be seen the messages provided as part of the ASSIST system are well estimated as indicated by the shape of the parameters and the speed contour plot shown in Figure 10. However, the length of congestion is both under and over reported during certain periods and is shifted forward in time. This may be due to the fact that the system may calculate congestion parameters based on previously measured data, causing a delay with computation and output. The time lag appears to be about 10-15 minutes. Overall, it seems that type 4 messages (not shown here) resembled actual traffic conditions closer than type 5 messages. The type 5 messages tend to underestimate congestion which could be caused by the use of the model. The model may dampen congestion estimates based on default values assumed in the model.

8

When sending high quality congestion messages, it is important to inform drivers about the Bottleneck Active 380-390

0 20

40 60

380-390

80 100 120 140 160

BN

15:00

16:00

17:00

18:00

19:00

20:00

congestion beforehand so they may anticipate the shock traveling upstream in a suitable manner. The majority of messages deployed by the ASSIST system to drivers indicated congestion downstream. Congestion was referenced to be downstream of a freeway exit to allow drivers to exit the freeway in anticipation of congested conditions. A total of 79 type 4 and 68 5 type messages were reported on June 27, 2002. Figure 10– June 27, 2002 Northbound A9 ASSIST Messages

Bottleneck Active 380-390

380-390

1 Congestion 2 Attention to Congestion

BN

15:00

16:00

17:00

18:00

9

19:00

20:00

Figure 11 – June 27, 2002 Northbound (VSL/ASSIST) With a bed of rich data resources, a further comparison was explored to investigate the relationship between when the VSL system indicated congestion to drivers as to that of the ASSIST system. Figure 11 shows the relationship for type 5 ASSIST messages as compared to the congestion indicated as was shown in Figure 8. The message sign locations are labeled on the y-axis. As shown, the comparison is quite similar however the message signs report congestion approximately 10-minutes earlier than the ASSIST system. It is determined that both systems well indicated drivers of congested conditions. FINAL REMARKS This paper describes a portion of a study that analyzed traffic conditions along a 16-km northbound section of Autobahn 9 between Munich and Nürnberg, Germany. This section of freeway houses several traffic flow surveillance and management tools used to better manage and improve traffic flow operations. This study used as its primary methodology transformed curves of cumulative count and time averaged velocity versus time to carefully and systematically diagnose bottleneck activations and deactivations. Further uses of time-space diagrams of speed and other parameters were also helpful in providing a better understanding of the dynamics of the variable speed limit and traveler information systems available along this corridor. In previous work [e.g., 3, 4] it has been shown that bottlenecks arose in a rather predictable manner in that flows increased above some level, queues were formed and propagated upstream, and demand reductions led to queue dissipation. To analyze possible bottleneck causal factors, other data sources obtained from this section of freeway included variable speed limit and message sign data, in vehicle navigation system data, floating car data, and weather data. These resources were used to aid with the triggers of bottleneck activation and determine there interaction with real traffic patterns. Several key observations and findings have arisen from this analysis. For example, it was found that different rules of the road apply to U.S. and German freeways. Autobahns are equipped with variable speed limit systems and drivers tend to use the median travel lane for passing only. Trucks are restricted from changing lanes at times and often Autobahns contain no regulatory speed limit. Control strategies such as ramp metering on freeway on-ramps are implemented on U.S. freeways and not Autobahns at this time. Further, it has been shown that average vehicle speeds are in compliance with that of the variable speed limit system located overhead on variable message signs. In addition, due to the availability of separate auto and truck sensor data, truck speeds were found to be on average 10-20 km/hr greater than the regulatory 80 km/hr speed limit and 12 km/hr greater just prior to bottleneck activation. The variable speed limit system was found to be related to actual traffic conditions and was active prior and during bottleneck activation. Speeds were generally found to be reduced after surges in decreasing flow and speed were observed. Vehicle warnings/restrictions indicated overhead on variable message signs were found to be active prior and during bottleneck activation. Congestion as indicated by the system was found to be consistent with actual traffic conditions. An examination of the BMW ASSIST data revealed that it provided a reasonable estimate of congestion parameters to drivers traveling along the study section of Autobahn 9.

10

Type 4 messages developed from raw data were found to more closely resemble actual traffic conditions than type 5 messages. This paper has sought to report on an empirical analysis of features of traffic dynamics and driver behavior on a German highway. Through this analysis, a comparison has been made between the behavior of German and U.S. drivers as they approach and pass through freeway bottlenecks. Research at this site is continuing this research is only an initial step toward understanding bottleneck behavior along freeways with different geometric features, traffic control, rules of the road, and driver behavior. Therefore, further analyses are being conducted at this site and other sites in Europe and the United States. ACKNOWLEDGEMENTS The authors acknowledge the support of Dr. Georg Lerner, Kristina Laffkas and the BMW Group. Thanks also to Steven Hansen and Nick Carey of Portland State University. Christian Mayr and Dr. Thomas Linder of Autobahndirektion Südbayern generously provided most of the data used in this analysis, and we are indebted to them for their assistance. REFERENCES [1] Cassidy, M.J., and J.R. Windover. Methodology for Assessing Dynamics of Freeway Traffic Flow. In Transportation Research Record 1484, TRB, National Research Council, Washington, D.C., 1997, pp. 73-79. [2]

Muñoz, J.C., and C.F. Daganzo. Fingerprinting Traffic from Static Freeway Sensors. Cooperative Transportation Dynamics, Vol. 1, 2002, pp. 1.1-1.11.

[3]

Cassidy, M.J., and R.L. Bertini. Some Traffic Features at Freeway Bottlenecks. Transportation Research, Vol. 33B, 1999, pp. 25-42.

[4]

Bertini, R.L., Hansen, S. and Bogenberger, K. Empirical Analysis of Traffic Sensor Data Surrounding a Bottleneck on a German Autobahn. Transportation Research Record: Journal of the Transportation Research Board, Washington, D.C., 2004. (In Press).

11