The condition of road surfaces can be easily identified using the accelerometer and gyro sensors of a smartphone. This has been demonstrated in testing performed by Gorski Consulting using an iPhone and multiple video cameras attached to a 2007 Buick Allure passenger car. Data from such testing is shown on the Road Data webpage of this Gorski Consulting website. The longitudinal and lateral motions of the test vehicle contained in this database were determined to be good indicators of the condition of a road surface. Factors that affect these values include the test vehicle’s speed and the suspension/body of the vehicle.
Qualities of a road segment are obtained by various municipal and state/provincial governments. Vehicles equipped with high-speed profilers are used to identify the physical layout of a road surface. These road segments are given a International Roughness Index (IRI) value. Unless one is a member of such agencies there is no practical way to obtain similar data. And government agencies are not willing to make this data publicly available. For those who are unable to obtain such data there is no objective way to question whether a road surface is adequate or needs repair except to conduct expensive laser scans for limited distances. The use of vehicle motion data provides that objective test, not focusing on the micro geometry of the surface, but focusing on what responses it creates in the motion of the test vehicle.
There are drawbacks to motion data. Motion data varies depending on the speed of the vehicle. It also likely varies from one vehicle to another. How much variance is related to vehicle differences is generally unknown because our testing has involved only one vehicle: a 2007 Buick Allure passenger car. It would be useful to examine the motions of vehicles that are very different from passenger cars.
Recently Gorski Consulting has gained the opportunity to study the motions of school buses. This is helpful because school buses are different in structure and suspension than a typical passenger car. Thus this enables the opportunity to study inter-vehicle differences in motion data. Two school buses were used in the testing:
- 2012 GMC 18-Passenger School Bus
- 2012 International 72-Passenger School Bus
These buses are shown in the photos below.
Two tests were performed with the GMC School Bus on March 4, 2021 in London, Ontario. In one test the bus was driven along a northbound route of White Oaks Road, Southdale Road, Wharncliffe Road and Western Road. In a second test the bus was driven northbound along White Oak Road, Southdale Road, Wellington Road and King Street. Two tests were also performed with the full-size, 2012 International School Bus on March 26, 2021. Both tests were performed eastbound from Pack Road, Bostwick Road, Exeter Road and White Oak Road.
In this article we will focus on the first test with the 2012 GMC 18-passenger school bus.
Previous testing was performed on Wharncliffe Road In London Ontario on March 31, 2014 using the 2007 Buick Allure passenger car. At that time several segments of the road surface were in disrepair. The results of the March, 2014 testing are posted on this Gorski Consulting website on the Road Data webpage. For convenience the data is reprinted in the following table.
The last two columns in the above table report the extent of lateral and longitudinal motion of the test vehicle in terms of radians per second. One radian is equal to 57.3 degrees. It may be recalled from previous discussions that the magnitude of vehicle motion can be evaluated according to three levels of severity:
- Rotation values up to 0.0200 rad/sec indicate a road surface that produces mild effects on the vehicle’s motion and therefore the surface is in good condition.
- Rotation values between 0.0200 and 0.0500 indicate a road surface that produces moderate levels of vehicle motion indicating that some portions the road segment could be substandard.
- Rotation values above 0.0500 indicate that the road segment contains major surface problems that could be a factor in the stability and safety of a travelling vehicle.
In the above table values in green indicate a low vehicle motion, below 0.0200 rad/sec and therefore a good quality road surface. The values in red indicate a high level of vehicle motion, above 0.0500 rad/sec and therefore a poor road surface. It can be seen that, as the test vehicle passed north of Duchess Ave, its motions became excessive and therefore the road surface was poor. Subsequent to the testing in 2014 the segments of Wharncliffe Road that contained the worse conditions were repaved. Thus at the time of the re-testing with the GMC School Bus in March, 2021, the road surface conditions were improved.
The testing on March 4, 2021 was conducted along the route shown in the following figure. The testing commenced near the intersection of White Oaks Road and Bradley Ave. Northbound travel along White Oaks Road was transferred for a brief distance onto Southdale Road before proceeding northward along Wharncliffe Road up to Oxford Street. Although the testing continued past Oxford we have limited the discussion up to Oxford.
The two tables below show the results of the March 4, 2021 testing over a period of 930 seconds (15.5 minutes). For reasons beyond our control the testing had to be performed during a morning rush hour. Thus in several instances the bus had to be brought to a stop, sometimes to wait for a red traffic signal. At other times the rush hour congestion caused the bus to stop in a line of stopped traffic. During these stoppages the recording indicated very low motions, as noted in the two tables. In contrast the testing in 2014 occurred in non-rush hour conditions and there was little interruption in the test vehicle’s travel.
Looking back at the testing from March 14, 2014 it was noted that the passenger car experienced a very large jolt as it crossed the north junction of the Thames River bridge. This disruption can be seen in the chart shown below. The chart shows the car’s travel over a time of 20 seconds. There is a smaller jolt in the longitudinal rotation at about 6-7 seconds into the chart which is probably the vehicle passing over the south junction of the bridge. Then there is the very large jolt at about 15 seconds which was caused when crossing the north junction of the bridge.
The above chart is broken down into greater detail in the next chart where we display the vehicle’s motion over a one-second time interval as it passes over the north junction of the bridge. The very large jolt occurs over a relatively short time and there is only one spike that approaches 2.4000 radians per second. Recall that the values being presented here are the rates of motion. In other words, how fast is that motion? So the value of 2.4000 radians per second means that, over a very short time, of perhaps less than a 10th of a second, the vehicle’s motion was 2.4000 radians per second. In fact, the reaction of the car occurred over a time of about 1/3 of the chart, or about 1/3 of a second and the longitudinal motion was generally in the range of 0.5000 radians per second – still very large when compared to the rest of the motion data.
The table below shows the data from the above chart in numerical form. Here we can clearly see what motion values were detected at each of the 30 samples of the 1 second time interval. We can see that only one sample (#1826.57) displayed a value of 2.3960 radians per second. The Standard Deviation value at the bottom of the table shows that, on average, the deviations were 0.5366 Longitudinal and 0.0942 rad/sec Lateral. So there was a single sample of a very high value that made the visual chart (above) seem more dramatic than it was. Yet, the rest of the data still suggests a very large reaction of the vehicle to the character of the road at the north bridge junction.
To see what caused this motion, the five photos below are frames taken from video during the March 31, 2014 testing. These views are looking through the windshield of the passenger car. They begin as the car approached the bridge at its south junction and carries on past the north junction.
There is nothing obvious in the above photos that could warn a driver of these large effects on the vehicle motion. This demonstrates a fact from previous studies that, in most situations, drivers are unable to detect many dangerous road surface features until it is too late to take any meaningful action.
Returning to the testing of March 4, 2021, the two charts below show the motions of the school bus as it travelled northbound through a similar location on Wharncliffe Road. Since this testing was preliminary only two video cameras were used; one showing the speedometer and one pointing forward through the windshield. Thus the precise surface feature that caused the motions could not be determined. In contrast 9 video cameras were used in the testing of March 31, 2014. Thus in 2014 there were several views that provided detailed information about the position of the test vehicle and the conditions of the surface that caused the motion.
The first chart shows the data as the bus passes underneath the CNR bridge and then reaches The Ridgeway crossroad. Our table shows that during this 30 seconds of travel the Longitudinal Rotation was 0.0449 and the Lateral was 0.0450 radians per second. The values are elevated but not into the red category (above 0.0500) that would suggest major road surface problems.
The next chart shows the motion in the vicinity of the crossing of the Thames River bridge. A large spike in the Lateral motion of over 0.4000 radians per second occurs right at the beginning of the chart, then there is a smaller spike at about 10-11 seconds followed by a larger, third spike around 16-17 seconds that rises just over 0.3000 radians per second. The last (third) spike is likely from riding over the north bridge junction. Although it is very large it is nowhere close the very large spike of 2.4000 radians per second that occurred in the 2014 testing.
When viewed in a graphic form, the data can give an exaggerated appearance of the vehicle motion when a single sample reports a very high value. When sampling rates are 30 or more per second, a single sample does not say much about the effect on the vehicle. Thus we need to look at a range or several samples to conclude that an effect on the vehicle motion is sufficient to be of concern.
Also there appears to be something unusual between the motions of the test vehicles in the two testing dates in the vicinity of the north bridge junction. In March 31, 1014 the data showed that the car experienced a very high longitudinal motion whereas in the March 4, 2021 data the Bus experienced a high lateral motion. These differences could related to repaving of the surface yet they remain puzzling. When using an iPhone to collect data it is possible to position the phone in different orientations. Also it is possible to mis-read the title of a column of data and what is meant by the author of the app. Thus it would not be too uncommon if the data in the March 31, 2014 data was misinterpreted such that the longitudinal and lateral motions were reversed. Our tests with the iPhone confirm that the interpretation of the March 4, 2021 is correct.
A quick check of the video during the 2014 testing indicates that the top of the iPhone was oriented to the left (toward the driver). In contrast, in 2021 the top of the iPhone was oriented toward the front of the vehicle. Thus the two phones were oriented with a 90 degrees difference. Thus what is reported as Longitudinal in one dataset will be lateral in the other, and vise versa. Reviewing the datasets it appears we made the correct adjustment for the differences in orientation of the iPhone, so the differences do not appear to be an error on our part. Yet the differences are so opposite that we will remain alert for possible explanations.
Meanwhile, the following 15 figures are frames taken from video during the testing of March 4, 2021. These views commence as the school bus is travelling under the CNR underpass and terminate as the bus is passing over the north junction of the Thames River bridge. These views provide some guidance about the conditions experienced during the March 2021 testing.
This preliminary testing with a 2012 GMC 18-passenger school bus has revealed that motion data can be generated that appears to be reliable in depicting the quality of a road surface. It is similar to that produced with a passenger car test vehicle in the sense that the motions are consistent with what would be expected. Travelling over a rougher section of road surface has produced higher levels of motion of the bus. When the bus has been brought to a stop we see that the motion becomes very minimal.
The only peculiar finding is with respect to the motions that occurred when the Bus travelled over the north junction of the Thames River bridge on Wharncliffe Road. In previous testing conducted ion March 31, 2014 a passenger car sustained very high levels of longitudinal motion when crossing this junction. In contrast the testing with the Bus produced a high lateral motion when crossing over the same junction. It is unclear if these differences occurred because of repaving of the surface between 2014 and 2021. It could also mean that the difference is related to the test vehicle differences.
This is the first test conducted with a school bus and it is expected that further testing will occur. We hope to discuss the results of our second test with the same school bus. Also two additional tests were performed with a full-size school bus on March 25, 2021 and hopefully there will be an opportunity to post the results in the near future.