Artificial Intelligence In Transportation Requires Clean and Unbiased Data
The fast-evolving areas of artificial intelligence (AI) hold great promise in road transportation as sensing of vehicle motion becomes incorporated in new vehicles and within roadway infrastructure. While there are a lot of big data out there that can be used to train the algorithms of this machine intelligence, some of it is unclean and biased. While sensing within vehicles and roadway infrastructure may be less susceptible to tampering, broad data-sets exist that have been developed from biased human judgment. Their existent is demonstrating the lack of past foresight in ensuring that this big data is valid and reliable for AI purposes.
A simple example of the AI developments that are taking shape is discussed in a research paper from 2016 entitled “Road crack detection using deep convolutional neural network”, by Lei Zhang et. al. and published in the 2106 IEEE International Conference. Basically images of a road surface were used to detect cracks in that surface for roadway maintenance purposes.
Data collection methods by Gorski Consulting of roadway surface characteristics have been discussed on this Gorski Consulting site since 2014 and a “Road Data” database exists on this website showing the results from the sensed motion of our test vehicle as it has travelled over many roads in Southwestern Ontario. Furthermore the data can also be matched to the videos from multiple video cameras that were attached to the test vehicle as the motion data was collected. Such a data-set is unique, clean and unbiased.
Examples of biased data comes from the vast numbers of police records of collision data. Such data is incomplete as little is being done to understand and recognize the large numbers of unreported incidents that are assumed to be of lower severity and therefore of minimal importance. Documentations by Gorski Consulting for the last nine years at a specific site (Clarke Road north of Fanshawe Park Road in London, Ontario) have involved photographic evidence of collisions and loss-of-control incidents that present a much clearer picture of the large numbers of collisions and incidents that are not reported in police data. At the specific site which includes a complex S-curve, between 80 and 90 percent incidents have been unreported in police data. While many of these incidents are minor, some are not and there is no information available in official records about how many significant incidents remain unreported. The involvement of placing blame for collisions provides further erosion of the police data as the motivation for documentation is not always focused on pure scientific purpose.
These issues demonstrate that the usefulness of the rapidly developing areas of AI are dependent on reliable and valid data that is not always available. Bad data is particularly damaging as it is difficult and time-consuming to detect the good from the bad. It now becomes more clear that it is of the utmost importance to examine how such data can be eradicated and collection methods improved.
Windshield Penetration Only Looks Horrific – But Raises An Important Safety Issue
The penetration of a car windshield by a flying piece of plywood looked horrific when presented in several photographs by the Ontario Provincial Police (OPP) on their Twitter account on January 2, 2019. The incident occurred on Hwy 410 just northeast of Toronto, Ontario. The OPP photos are shown below.
In actuality the passengers were not “lucky to be alive” as claimed on the OPP Twitter account. The piece of wood has come to rest while still embedded in the windshield. This means that “just before” being embedded it was travelling very slowly. Progressing slightly further back in time it was moving faster with respect to the vehicle but it was also not penetrating into the interior as much as when it was embedded at final rest. So, no, in this specific incident the occupants were not “lucky to be alive” but experienced the reported minor injuries that would be expected.
On the other hand, the OPP Twitter posting raised an important point about the danger of more serious incidents. While you are travelling at 110 km/h in one direction a piece of debris from an opposing vehicle may be travelling at 110 km/h in the opposing direction. That is a closing speed of 220 km/h. That debris would be coming at you at about 61 metres every second. If you consider a reaction time of about 2 seconds to initiate a response and then consider that your vehicle cannot steer away or brake “on a dime”, you are very unlikely to avoid a piece of flying debris like this unless you have a great deal of prior warning. In fact in almost all incidents drivers cannot initiate a successful evasive action and avoiding such an impact is very much a case of luck.
Even a piece of debris that weighs very little (i.e. is of a small mass) becomes dangerous because of the speed (velocity in the opposite direction toward you) at which it is closing toward you. So even very small objects, when they are travelling very quickly toward you can be lethal. Broad objects like a piece of plywood may be less dangerous when they strike you with the broad, flat surface. But turning that plywood around 90 degrees so that the edge is coming at you will make a world of difference with respect to your injury and survival. No seat-belts or air bags will save you.
Modern windshields are actually quite resistant to penetration as they are formed of two pieces of tempered glass with a thick plastic layer in between. It is that plastic layer that keeps the windshield intact even though the glass may become completely destroyed. In severe head impacts during severe frontal collisions, unrestrained occupants have been known to tear that plastic layer to some degree but complete head penetration through that layer would be very rare. It is those small but solid pieces of debris (something like the size of a bowling ball) that can be extremely dangerous when they contain ragged edges that can pierce through the plastic and have enough mass to apply a considerable “point load” to a small part of the windshield. Obviously anything more massive increases the danger.
Considering the danger of possible debris being projected from one side of a divided highway onto the other is would be advisable to consider driving in the right lane of such a multi-lane highway when such a choice is possible. While these incidents are rare, they are difficult to avoid and can be lethal.
Headlight Performance & Cost May Surprise You
An article in the November 29, 2018 Status Report of the Insurance Institute for Highway Safety (IIHS) may be surprising with respect to the performance and costs of new headlights. The IIHS studies show the poor performance of headlights on some new car models from a variety of manufacturers. In a graphic (shown below) the IIHS is pleased to show that the performance of headlights has been improving between 2016 and 2018 model years.
IIHS study showing improvement in headlight performance.
Yet, further in the article, the IIHS demonstrate the massive cost for the improvement of this performance, as shown below.
Would you be happy to pay $4,000.00 for a replacement headlight?
The Status Report article confirmed “When we did an initial survey of prices last year for 2018 models, Ford was charging $4,555,00 for a Lincoln Continental headlight, the most expensive one in our survey”.
In contrast in 2017 a replacement of a headlight bulb for a 2007 Buick Allure at Canadian Tire resulted in a total cost of $40.44, including taxes.
The problem with these facts is that improved headlight performance may not necessarily improve roadway safety if costs are incredibly high. Many vehicle owners might consider driving without replacing a non-functioning headlight and take their chances at receiving a traffic ticket at a much lower cost. Would we like to meet one of these vehicles on the highway of a dark and rainy night?
Guardrail System Installed & Functioned Properly Yet Fatality Occurred
A properly-installed guardrail was struck on Southminster Bourne Road southwest of London, Ontario on December 26, 2018 yet a fatality still occurred. What caused this fatal result to the 27-year old female driver is unknown. No photographs were made available by the local news media showing the damage to the vehicle and police have yet to provide an explanation why the fatality occurred.
Initially local news media reported that the vehicle struck a “bridge barrier”. This was not helpful because this did not provide a clear indication of what was struck. Subsequently the site was examined by Gorski Consulting on the morning of December 27, 2018 and this provided a clearer indication of what occurred.
The photo below shows the collision site which is on the north side of Highway 401. Because of the meandering character of Southminster Bourne Road the actual travel direction of the vehicle at the accident site was eastward even though the road crosses Highway 401 which runs from north-east to south-west.
The case vehicle was eastbound on Southminster Bourne Road and approached the left curve shown in the background. The overpass to Highway 401 can be seen in the background.
Some posts are visible along the right side of the road and these would normally be accompanied by a cable barrier. Even though the cable was missing it was not a factor in this collision as it was not involved the collision events.
The physical evidence leading to the collision was classic of what has been observed over several decades of hundreds of similar cases. As the vehicle entered the left curve its right side tires travelled off of the paved surface and onto the gravel shoulder. This caused a counter-clockwise rotation causing the vehicle to begin sliding sideways, leading with its right side, as it crossed the roadway centre-line and then impacted the guardrail on the opposite side of the road just before reaching a small bridge.
The photo below shows the right edge of the road and here it can be seen that a substantial width of additional pavement exists to the outside of the white, painted, edge line of the lane. In the past such an additional amount of pavement did not exist along most rural highways and this led to many loss-of-control events. So because of this additional width of pavement this road segment could be classified as well-designed and equipped.
A substantial width of additional pavement existed to the right of the white painted edge line and this is an additional safety feature that would normally reduce the number of loss-of-control rotations and subsequent collisions.
As typical there was a moderate super-elevation in the eastbound lane leading through the curve. Such a super-elevation involves raising the right edge of the lane so that a lateral down-slope in created toward the centre-line. This “banking’ feature creates an additional lateral force that helps to pull the vehicle to the left and makes it easier for a driver to pass through the left curve. Often the right shoulder at such a super-elevation is not raised like the pavement but is sloped in the opposite direction. This causes a difficulty when the right side tires of a vehicle travel from the paved surface which is super-elevated, to a gravel surface which is not. In addition the gravel produces more drag on those right side tires causing an initial, slight, pull to the right which is sometimes counteracted by a driver steering to the left. Thus these are some of the complicated forces that exist which lead to an eventual counter-clockwise rotation.
The photo below shows the area where the vehicle left the right shoulder and slid across to impact of the guardrail on the opposite side of the road. In the pavement one can see the curved “yaw” marks that were generated as the vehicle proceeded through its counter-clockwise rotation. Typically the most visible tire marks are produced by the two right side tires because the weight of the vehicle is transferred to them during the rotation and deceleration. As the rotation progresses the observer then begins to see the markings from the two left side tires which are usually fainter.
The obvious sign that the vehicle entered into a counter-clockwise rotation is the visible “yaw” marks on the pavement leading to the guardrail impact on the opposite side of the road in the background.
With the advent of event data recorders it is possible to determine the travel speed of the vehicle before the vehicle entered into the yaw. Because such speed data is often related to the rotation speed of the tires investigators must be careful in properly interpreting such data. For example it is possible for a vehicle sliding sideways at 100 km/h to register speed data that indicates that the vehicle is stopped.
Investigators have always had the ability to measure the curvature of yaw marks and determine the vehicle’s speed. Also the angle of lateral striations within the marks can also indicate whether the vehicle was being braked or was being accelerated.
The photo below shows that the yaw marks converge meaning that the vehicle has rotated to a position where it is sliding sideways, leading with its right side as it approaches its impact with the guardrail.
This photo shows that the yaw marks from the rear tires converge before reaching the guardrail and this means that the vehicle has reached a point where it was sliding sideways, leading with its right side, just as it approached the guardrail.
The photo below shows how the guardrail was displaced as a result of the impact. The anchorage posts became separated and deformed. All this is precisely what the system was designed to do. Through its deflection the system provides more time for the striking vehicle to reduce its speed. The destruction of the rail and posts is designed to remove the kinetic energy that exists in the striking vehicle as a result of its mass and velocity.
The guardrail was struck just before the vehicle reached the bridge. The rail was displaced as it is designed to do.
The photo below shows that the guardrail was properly anchored to the concrete bridge abutment and there were additional anchorage posts as the rail approached the concrete abutment. This stiffening of the rail is what is supposed to deflect the striking vehicle away from the immovable and dangerous concrete abutment. In olden days guardrails were poorly constructed such that they were not properly stiffened nor anchored to bridge abutments. As a a result vehicles that struck the railing were often chanelled into an impact with the abutment resulting in fatal injuries. Such an incident occurred on Highway 401 at Dodd’s Creek, just west of London in the late 1970s resulting in an inquest. As a result of that inquest the Ontario Ministry of Transportation began a review of its guardrail anchorages and it was mandated that all guardrails should be anchored in the manner shown in these photos.
The guardrail system is purposely designed with additional posts near its anchorage at the concrete bridge abutment. This is what re-directs the vehicle away from the dangerous, immovable abutment.
The backside of the guardrail is shown in the photo below and it provides a further view of the stiffening of the rail as it approaches the concrete bridge abutment.
This backside view of the guardrail shows how it has been stiffened by an increase in the number of anchorage posts as it approaches the bridge abutment. Again, this stiffening is meant to deflect a striking vehicle away from the immovable concrete abutment.
While the vehicle was deflected during the initial impact it also continued to rotate counter-clockwise and therefore its rear end struck the railing of the bridge further to the east as shown in the photo below.
View of the metal bridge railing where the rear of the rotating vehicle made contact after its initial contact with the guardrail.
The presence of this additional, secondary contact may explain why the local news media indicated that the vehicle struck a “bridge barrier”. While it is true that the railing was struck this secondary contact was of minimal consequence as the main threat was from the initial impact with the guardrail.
The physical evidence showed that after the initial and secondary contacts at the north side of the road the vehicle was deflected back toward the south where it made an additional, minor contact with the bridge railing and then settled to rest near the east end of the bridge, as shown in this photo.
The official reports of the incident indicated that it was the vehicle driver who sustained fatal injuries while a right front passenger was transported to hospital with serious injuries. This is not what one would expect from this collision. Since the direct impact occurred at the right front of the vehicle it was the right front occupant who should have been in greatest danger of sustaining the higher severity of injury. The reported driver of the vehicle should have been located at the far side (left) of the vehicle interior when the vehicle struck the guardrail and therefore it would be expected that she should sustain less injury than the right front occupant.
As stated earlier, regardless of where the occupants were positioned fatal injuries would not be expected to occur. The very purpose of a guardrail is to reduce the severity of injury and the deflection of the vehicle that occurred should have provided both occupants with a good opportunity to survive this impact. This is another example where the public needs to be better informed about these incidents. The death of a person cannot be a private matter when the reason why that death occurred is in question. There could be factors that could explain why the fatality occurred but an explanation should be made publicly by those who conducted the official investigation.
Highway 401 Westbound Traffic Volume Completed For Kenesserie Road Site
Calculation of traffic volumes have now been completed from the videotaping session on Wednesday, December 19, 2018. These calculations have been added to the table of the other sites and the new table is shown below.
The results show that the Wednesday videotaping session at Kenesserie Road resulted in heavy truck volume of almost 48%. This is consistent with the other weekday (Monday) session at Graham Road where the heavy truck percentage was 48.6%.
Two heavy truck collisions occurred this month with the recently-installed cable barrier. In both instances the cable barrier allowed the trucks to cross the median and enter into the opposing lanes of traffic. The photo below shows the cable barrier from an eastward view on the overpass at Kenesserie Road. It can be seen that the cable barrier exists only the the eastbound side of the median. Thus westbound vehicles are allowed to travel through the median and do not reach the barrier until they are only a couple of metres away from the eastbound lanes.
View looking eastward from the Kenesserie Road overpass on Highway 401. The new High Tension Cable Median Barrier is shown positioned along the north edge of the eastbound lanes. Videotaping was conducted of the westbound traffic.
Our experience from the 1980s has involved the detailed documentation of the travel paths of vehicles that passed through the median and were involved in median crossover collisions. The general public may not understand that, even though the median slope, width and surface are engineered to certain standards, the expectation that vehicles will travel through that median in a controlled and predictable manner is unreasonable. Our detailed measurements of the physical evidence along with our reconstruction of vehicle motions and paths has demonstrated that vehicle motion though such medians is chaotic and unpredictable. Thus impacts of the cable barrier on the far side of the median would also be expected to be chaotic and unpredictable. Since the heavy truck traffic volume on weekdays is likely to be approaching 50% of the total traffic many of these heavy trucks will travel through the median in an unpredictable trajectory and their manner of engagement with the cable barrier cannot be guaranteed.
It remains to be seen how the new cable barrier will perform as more impacts by heavy trucks are documented. However the crucial issue is that these incidents need to be properly and fully documented and that documentation needs to be made available to the general public who use this highway.
You must be logged in to post a comment.