Session 1: Field Operation Tests on the Sangubashi Section of Metropolitan Expressway No.4

I would like to begin my presentation today with a brief introduction to the way in which AHSRA has progressed with system construction to date, after which I will discuss the results of tests carried out on the Sangubashi section of Metropolitan Expressway No. 4, as well as looking briefly at the results of tests in other areas.

AHSRA commenced the project by defining the services to be offered by the system. From there, we progressed to formulating system concepts, clarifying requirements for element technologies, and designing and developing the system. We presented our system concepts in Japan and overseas, after which we made necessary modifications and conducted field operation tests. At present we are moving towards practical implementation of the system. (Figure 1)

Field Operation tests have been carried out in seven locations around the country. Tests in six of these locations were completed by fiscal 2003, but the seventh location, the Sangubashi section of the Shinjuku Line of Metropolitan Expressway No. 4, presented us with severe difficulties in terms of setting up the equipment, and tests therefore had to wait until fiscal 2004. I will be providing an overview of these tests and discussing the results in my presentation today. (Figure 2)

3. Outline of Field Operation Tests on Sangubashi Section of Metropolitan Expressway No. 4 (Shinjuku Line)

As shown in Figure 3, the test area on the Sangubashi section of the Shinjuku Line on Metropolitan Expressway No. 4 features a left curve of a radius of 88 m, an extremely sharp curve, running from the Shinjuku Interchange towards the center of the city. According to accident statistics, 140 accidents resulting in damage to vehicles and road infrastructure occurred on this section of road in fiscal 2002. Naturally, a variety of safety measures have already been implemented on the curve, but accidents are still occurring, and new measures are therefore called for.

We conducted AHS trials on this section of road. The services we tested were the provision of information on obstacles ahead and warnings on the presence of a curve. Trials were conducted on these services in combination.

When there is an obstacle around the curve, the system warns the driver before the vehicle enters the curve. The warning states “Traffic congestion (stationary vehicles) beyond curve. Drive carefully.” If the driver is approaching too fast for a curve of that radius, the system announces “Warning, curve approaching” and “Slow down.” When there is no obstacle around the curve, the system informs the driver of the presence of the curve, and provides warnings on speed as required. (Figure 4)

To give you some idea of what this is like in practice, I would like to show you an example of the type of services I am discussing. This photograph was taken on another curve on the Shinjuku Ramp leading in to the Sangubashi section. The service commences when the vehicle rounds the curve. (Figure 5: Voice messages “Slow-moving vehicles beyond curve. Drive carefully.” “Slow down! Road curves left” “Reduce speed”)

There was not a large volume of traffic on the road at the time when this photograph was taken, but there was an area of traffic congestion around the curve, and this demonstrates how useful it is to drivers to be provided with information before entering a curve.

(3) Overview of test system
To enable the realization of these services, we established infrastructure on the Sangubashi curve. We positioned four infrared sensors to enable detection of traffic congestion, stationary vehicles and slow-moving vehicles over the entire curve. The information gathered by these sensors was provided to drivers at an information transmission point. (Figure 6)

(4) Categories for verification
Turning to the main items we studied in the test area, first, we conducted a study of dangerous situations on the road. We analyzed the types of dangerous situations arising on this curve, and attempted to determine how AHS could be effective in these situations. Second, we investigated the functioning of the system overall, verifying the performance of the infrastructure, the sensors, and the communications system. Third, we had drivers do test drives, after which we gave them questionnaires (Figure 7). These were the major categories of our study.

< 1> Existence of “hidden” accidents
I will commence my discussion of the results of our study with the results of our survey of dangerous situations. First, our study showed that a very large number of hidden accidents occur on this curve. Our observations showed that 30 accidents occurred in the four-week test period (Figure 8). Of these, 12 were reported, which means that more accidents are hidden than appear in the statistics. If minor accidents like unreported bumps and scrapes are included in the statistics, we see that there are actually many more accidents on this section of road than is initially apparent. A variety of safety measures are therefore necessary.

A breakdown of the accidents occurring during the test period by type shows four rear-end collisions and 26 single-vehicle collisions with the side walls.

Figure 8

I would like to show you an example of one of these accidents. These are four images showing the entire curve, taken from cameras set up on the roofs of nearby buildings. The photograph in the upper left shows the most upstream point of the curve, from the straight section. The upper right shows a point two-quarters of the way into the curve. The lower left shows a point three-quarters into the curve, and the final photograph shows the exit of the curve.

In the photograph on the lower left a single-vehicle accident occurs, and the vehicle becomes stationary in a position that cannot be seen from upstream. As a result, the next vehicle around the curve hits the rear of that vehicle and they both become stationary. After a while, a truck comes around the corner without adjusting its speed because the road appears clear, and is involved in a serious rear-end collision with the stationary vehicles. We see here how vehicles become stationary around the curve where they cannot be seen from upstream, and why there are therefore numerous rear-end collisions on curves.

<2> Accidents caused by obstacles ahead
Let's look more closely at the details of these accidents (Figure 9). We have classified 11 of the 30 accidents occurring during the test period as accidents caused by obstacles ahead of the vehicle. These may be collisions with the rearmost congestion which the driver was unaware of, or collisions with the side wall caused by trying to avoid this type of accident.

There were also numerous secondary collisions in this category. In this type of accident a vehicle might take the curve too fast and be involved in a single vehicle accident, after which a following vehicle will be involved in a rear-end collision with it, or be involved in an accident trying to avoid such a collision. These accidents included, 11 of the accidents occurring during the test period were caused by obstacles ahead, which represents a fairly high percentage. We believe that the provision of information to drivers would be highly effective in the case of this type of accident.

We will look at secondary collisions in a little more detail. In three cases we observed, the lead vehicle was involved in a single-vehicle collision, after which the first, second, third and fourth following vehicles were either involved in rear-end collisions or accidents while attempting to avoid rear-end collisions. These secondary collisions occurred anywhere from 10 seconds to 17 minutes after the initial accident.

The areas shaded in blue in Figure 9 show accidents in which it would have been possible to provide information on the existence of an obstacle ahead at the information transmission point. If the vehicles had been fitted with receivers, they would have received that information. We can therefore indicate that the ability to receive information from the AHS would be useful in the case of secondary collisions and collisions with the rearmost congestion.

Figure 9

< 3> Accidents caused by excessive speed
There were also accidents caused by excessive speed. The purple section of the graph in Figure 10 shows the distribution of speeds at entry to the curve among vehicles involved in accidents caused by obstacles ahead. The blue sections show the distribution of speeds at entry to the curve among vehicles involved in collisions with the side walls caused by excessive speed. The blue sections show a spread of fairly high speeds, from 55 to 85 km/h. The design speed in this section of road is 50 km/h. In this case also, we believe that a suitable service targeting vehicle speed would be effective in reducing accidents.

Figure 10

< 4> Analysis of dangerous behavior (Near misses)
During the test period, there were also incidents which we classify as “near misses,” which did not eventuate in accidents. Our analysis of the test data focused on vehicles entering the curve at 40 km/h and over, because the Metropolitan Expressway Public Corporation classifies situations in which vehicles travel at speeds of under 40 km/h as congestion. (Figure 11)

On average, 200 vehicles per day entered the curve at 40 km/h or above when there was an obstacle ahead. Study of the amount of vehicles braking suddenly in this situation shows the very high proportion of 15%. It is not clear whether or not drivers experience a shock from these incidents, but it does represent dangerous behavior.

When there is no obstacle around the curve only 3.7% of drivers entering at 40 km/h or above brake suddenly. In other words, the proportion braking suddenly is four times higher when there is a forward obstacle. Taking this also into consideration, we can see the very high value of the provision of information on obstacles ahead to drivers.

Figure 11

<5> Estimate of effects of provision of information on obstacles ahead
Let's look at how AHS would function in the case of accidents of this type. In Figure 12 we see a graph of the vehicle trajectory in the accident we previously saw on video. This graph of vehicle trajectory was generated by sensor readings of vehicle position and speed. The horizontal axis shows time, and the vertical axis shows the distance the vehicle travels along the road, with the information transmission point as zero.

Sensors were located from 300 m to 450 m, and we were therefore able to track the trajectory of the vehicle over this distance. At first, the vehicle's trajectory is smooth. At a certain point, the vehicle is involved in a single-vehicle accident. The graph shows it becoming stationary, and here we see a truck collide with the vehicle in a secondary collision.

If we work backwards from 51.7 km/h, the speed at which the second vehicle involved in the accident, shown here as pink, entered the curve, we see that it could have received information concerning the previous accident ahead when it passed the information transmission point. The sensors have registered the accident, and are transmitting information.

Figure 12

We worked back in the same way for each of the accidents caused by obstacles ahead to determine whether or not the system was transmitting information on obstacles as the vehicles passed the information transmission point in each case. We determined that in 10 of these 11 accidents the vehicles would have received information on the presence of slow-moving vehicles, stationary vehicles or the rearmost congestion around the curve as they passed the beacon (Figure 13). In the remaining case, a rear-end collision occurred 10 seconds after the initial accident. The events in this sequence were too close together for the presentation of information; there was still no information on the original event when the second vehicle passed the beacon. Other measures are possible to correct this, such as positioning another beacon directly in front of the curve.

Figure 13

< 1> Road sensor monitoring system
I'll move on to discuss system verification. First, we studied the performance of the road sensors (Figure 14). Four infrared sensors were positioned on the test section of road. Cameras were placed on the roofs of nearby buildings to serve as a reference, enabling us to determine visually the kind of events that had occurred. The records from the sensors were studied in comparison with these visual images to enable sensor performance to be evaluated.

Looking at the main results, we found that over the four-week test period, 31 stationary vehicles were recorded. 19 of these vehicles became stationary after accidents. That is, 19 of these make up part of the 30 accidents I mentioned earlier. Another 12 vehicles were temporarily stationary due to road works and the like, making a total of 31. Of these, the sensors detected 30.

The case not detected by the sensors was a motorcycle which had fallen over. The infrared sensors registered a heat source, but because the motorcycle had fallen over, the position of the heat source was obscured, and the sensors could not accurately detect it. However, the performance of the sensors did clear our initial target of 96%.

We also verified sensor performance in random sampling of rearmost congestion. Time lags occurred to a greater or lesser degree, but sensors detected all the incidents of traffic congestion.

The system did register incorrect data, sometimes indicating the presence of rearmost congestion late at night. Close investigation of the cause showed that during road work the system had registered large stationary road work vehicles, the movements of road repair workers or the heat from flares as indicating rearmost congestion. This problem should be eliminated by modifying operation of the system during road work.

Figure 14

<2> DSRC performance
Coverage of the area, the position of the control point and the service zero point were studied to determine communications performance. The accuracy of the system was within plus or minus five meters, meeting our initial target. In addition, we ran numerous tests using test vehicles, and verified that the services were entirely functional in each case. (Figure 15)

Figure 15

These tests were conducted on an elevated Metropolitan Expressway, with ordinary roads running in the same direction below. Tests showed that leakage of the communications carrier waves on the expressway caused services to commence inappropriately in test vehicles running on the ordinary roads. We will have to study changing the position or adjusting the angle of the equipment to solve this problem, and in the future we will also need to investigate adding a service lane judgment function to the on-board unit.

We also studied the degree of communications shadowing. This occurs, for example, on a two-lane road when there is a truck in one lane and a passenger vehicle beyond it in the other, and the communications carrier wave is blocked by the truck and does not reach the other vehicle. We confirmed that shadowing does not occur with an antenna height of 8 m.

We had thirty test subjects drive AHS-equipped vehicles on the test section, after which we sought their opinion on the services. All of the subjects indicated that they understood the information the system presented to them. They believed that being presented with this information in advance was a clear aid to safe driving. (Figure 16)

(1) Effectiveness of provision of information on forward obstacles in reducing speed

I'd like now to look briefly at the results of tests conducted in areas other than Sangubashi. On a section of an expressway in Maitani and at the Nagoya Nishi Junction, we used message signs to present information gathered by the AHS to drivers. We studied the reduction in speed at entry to the curves when this information was presented and when it was not presented. The results showed an average speed reduction of six or seven km/h. To ensure fairness, we sampled vehicles driving in unimpeded circumstances with five seconds of headway, and compared speeds at entry to the curve. (Figure 17)

We believe that the ultimate criterion for evaluation of this system will be the extent to which it reduces accidents. We therefore did a cumulative study of the total number of accidents occurring on the section of expressway in Maitani 18 months before and after the introduction of the message sign. (Figure 18)

Looking at the total number of accidents, we see 47 before the placement of the signboard and 26 after, a reduction of almost one-half. Other safety measures have also contributed to the achievement of this result. However, the provision of information was the main measure employed here, and we believe the results demonstrate its effectiveness.

Allow me to sum up my presentation today. First, a study of close calls on the test section of expressway clarified the actual status of accidents on that section. When there are obstacles ahead around the curve, in addition to accidents, there are also a high proportion of incidents of sudden braking, and the provision of information prior to vehicles entering the curve will be effective in dealing with this situation. Analysis of actual accidents confirmed that if the vehicles involved had been fitted with receivers, information on the presence of forward obstacles could have been transmitted to them.

System verification showed that the system as designed for a section of road with an uninterrupted flow of traffic was able to meet or exceed performance targets in detecting traffic conditions and transmitting information to vehicles. We believe that this level of performance enables the system to be put into practical use. Finally, results from other areas show that the provision of information on obstacles ahead is effective in reducing the number of accidents and in reducing the speed at which vehicles enter curves.

On the basis of these results, we are currently moving towards the provision of AHS information to drivers, beginning with message signs and existing information provision systems.

2004 AHS Symposium

Session 1:

“Field Operation Tests on the Sangubashi Section of Metropolitan Expressway No. 4”