How to verify trust: the Caltrans results

By Joseph Averkamp

Start counting: 11 vehicles per minute travelling at highway speeds – that’s one every five seconds. How many people were in it? Are you sure? Here comes another one. The cars just keep on coming, so don’t blink!

“While the study isn’t the end of the story, it illustrates a path forward to a safer, less congested roadway where traffic flows smoothly.” — Joseph Averkamp, senior director, technology, policy and technical strategy

That’s the challenge police officers take up when they enforce your High Occupancy Vehicle (HOV) lanes or High Occupancy Toll (HOT) lanes. They can do it, but they’ll be accurate about 36 percent of the time. But what about issuing citations to violators? That will be difficult to do manually when you have approximately 78 violations per hour.

These numbers came from a test of Xerox technology that counts passengers in vehicles as they travelled at regular highway speed. We conducted these tests with Caltrans on a section of Interstate 5’s HOV lane in Orange County, California. This testing showed how transportation agencies can verify that their HOV and High Occupancy Toll (HOT) lanes are used as intended.

In addition to testing the Vehicle Passenger Detection System’s accuracy, our efforts also paved the way to the Best of ITS Award. This award was presented as part of Infrastructure of Things category, and we were selected over ten other finalists.

Subscribe to email updates and receive the latest transportation industry trends and tips.

This press release tells you about the award, and this blog article tells you why transportation agencies like Caltrans need an automated tool that counts passengers in vehicles. Here is what we learned from our test on a section of Interstate 5 in Orange County, California:

  • 671 vehicles per hour in the HOV lane. That’s 11 vehicles per minute, or one vehicle every five seconds. Being able to look into the vehicle and make a snap judgment as to whether or not the vehicle is HOV-qualified is challenging. The vehicles just never stop coming.
  • Approximately 78 violators per hour. Even accounting for Low Emission Vehicles, which are allowed in the HOV lane, the violation rate was 11.65 percent. An enforcement officer can manage three to four enforcement actions an hour if a roadside stop is required. An agency could not place enough officers on the roadside to cite 78 vehicles in an hour. The impacts to safety and congestion would be intolerable, and the cost would be astronomical. Roadside enforcement for numbers at this level is not possible.
  • Roadside observers were 36 percent accurate; the Xerox Vehicle Passenger Detection System was 95 percent accurate in determining whether a vehicle was a Single Occupant Vehicle (SOV) or an HOV. Moreover, our system captures high quality images which can be used in two ways.
    1. Allows manual reviewers to examine the images in the safety of their office, and reject those where the system has mistakenly declared the SOV/HOV status. This allows the accuracy results to be improved beyond 95-96 percent range.
    2. The images can become an evidence package that captures the results for future use.

Unfortunately, for humans using their eyeballs on the roadside, there is no opportunity for a second review or to capture an evidence package.

The trial was conducted over three days in January, 2015. The periods of assessment were the morning and afternoon rush hours,  6 a.m. to 9 a.m. and 3 p.m. to 6 p.m. each day. The intent was to compare accuracy of human observers and our Vehicle Passenger Detection System.

At the conclusion of the study, Joe Rouse, managed lane director for Caltrans, said he was “stunned” by the numbers, and believes that the Xerox system would be an effective tool in policing HOV lanes.  “For us, it all comes down to keep traffic moving by making best use of the infrastructure we have,” he said.

Here’s a look at our test results:

Statistic Measure Comments
Total Vehicles Reviewed 12,073
Violation Rate 11.65% 17.4% adjusted down for low emission vehicles.
Total Violations 1,406 11.65% * 12,073
Number of Hours During Analysis Period 18 3 days for 6 hours each day
Number of Vehicles Seen Per Hour 671 11 vehicles per minute or one vehicle every 5.4 seconds
Violators Per Hour 78
Single Occupant Vehicle (SOV) Detection Accuracy Xerox VPDS: 95.3%

Human Roadside Observers: 35.6%

 

This presents a new way forward for HOV/HOT management. In my previous blog post, we discussed the key challenges of equity, safety, congestion and reliability. This study demonstrates that these issues can be managed, and while the study isn’t the end of the story, it illustrates a path forward to a safer, less congested roadway where traffic flows smoothly. Xerox is proud to be contribute to advancing the state of the art for transportation  systems and we’re delighted to accept the Best of ITS Award 2016The results are compelling and the study shows the challenges to enforcing the rules on HOV/HOT lanes, and demonstrates a solution to the problem.

20 thoughts on “How to verify trust: the Caltrans results

  1. Jackie Messina August 5, 2016 - Reply

    Just like a copier . The main page is scan first before print . The road (Hwy ) can be scan . Recorded travel slowed then print of Violators . Mail a letter of warning showing the Violation. Car in system comes up again second warning sent . Once more you will recieve ticket or no pass . Less jams and a few picker fingers keeps the traffic rolling. ?
    AcE

  2. Why two warnings? You think they didn’t know they were breaking the law?

    Zero tolerance. There are signs all over the place telling people the penalty, they just don’t care because they know they’re not going to get caught.

    • Agreed. This blog likely only exacerbates the problem, since it illustrates that for the moment, at least, being ticketed for improper HOV use is statistically unlikely.

      • Joe Averkamp August 11, 2016 - Reply

        Martin and Chris, Zero tolerance can be a good approach, but the challenge is how do you identify and enforce the violations? In the study there were 78 violations per hour. To enforce that many violations would require several police vehicles along the roadside, which would, by itself, cause congestion. This system is intended to be more accurate than humans, be less intrusive in that there will be less rubber-necking than there is with a police vehicle, and be safer—officers conducting a stop are exposed to tremendous risk just due to traffic. With the system, the officer does not need to stop the violator vehicle. Agencies can pursue a zero tolerance approach— this tool makes it easier and safer to identify violators.

  3. I am not in L.A. area very often. When I am I love using the HOV lanes. However, when I am unable to, adding those violators to the other lanes will cause even more congestion. Or will they continue to use the HOV lanes and just add to the city,county,state revenue? Would doing away with the HOV lanes lessen congestion? How about doing that study? I understand the attempt to make people carpool for work, but this is California and that is not going to happen.

    • The intent is to encourage people to carpool, and one way to do that is to provide an incentive (HOV lane). Of course, many people will not carpool and will greedily attempt to reap the incentive anyway. The state could sure use the revenue from such people.

      • Joe Averkamp August 11, 2016 - Reply

        Chris, I agree with your assessment. The HOV lanes are intended to increase carpooling and reduce congestion. The Vehicle Passenger Detection System doesn’t really create violators, it simply is a tool to help state highway officials and law enforcement identify the violators. Studies have been conducted that show that properly managed these lanes can improve throughput of people. This now takes the form of High Occupancy Vehicle lanes or High Occupancy Toll lanes—where you can ride as a single occupant driver, but you must pay a toll. This has the benefit of raising revenue for additional infrastructure to be built and to managing demand.

      • Driving Dave August 13, 2016 - Reply

        Study after study show that HOV lanes do NOT encourage people to carpool. People carpool when it’s convenient and don’t when it’s not. If we were really concerned about congestion HOV lanes would be abolished.

        • Joe Averkamp August 15, 2016 - Reply

          State agencies are reluctant to give up on HOV lanes. The challenge is that when we give away valuable goods and services for free, the demand simply overwhelms the supply. In the case of road systems, highly travelled roads at rush hour are a valuable commodity, and if we give them away at no charge, the users will ALL want to use the road, creating congestion. HOV lanes were an attempt to manage demand by requiring people to take some action that has a cost–that is, lining up other riders. Where most agencies will migrate is to High Occupancy Toll lanes—-the user can either add riders or pay the toll. This has the twin benefits of managing demand (not everyone is willing to pay) and it raises funds for road maintenance and expansion. The underlying challenge is that geography and funding prevent us from building as much capacity as we could use if the price is free—so we need to establish a price for use, in just the same way airlines, trains, and buses do. The road isn’t actually free.

  4. Are they able to see a child in a carseat in the middle seat in the back of a midsize suv? Or are they only looking at front seat passengers? Cause kids aren’t supposed to sit up front until something like age 12 nowadays.

    • Joe Averkamp August 11, 2016 - Reply

      Amanda, that is a great question. During the trial, we were able to spot children in car seats with good success. However, since there will be errors, our recommendation to states and agencies implementing the system is to add a manual image review step after the automated score is developed to reduce mistakes that the system may make. And also, when there is doubt about the result, err on the side of forgiving the violation—agencies will be able pursue their own strategies for managing single occupant drivers in the HOV lane. We are trying to provide a tool that makes the task safer and more equitable. On a lane with 15,000 vehicles per day (or so), the vast majority are law-abiding (in our studies 80-90%), so since they follow the rules, the equitable approach is to encourage others to also follow the rules.

  5. Will probably add $500 million in revenue a year and increase traffic safety by aproximately 0%. How again does any of this increase public safety?

  6. If a cop pulls someone over and realizes a mistake was made, no HOV ticket gets issued.

    If an algorithm makes a mistake, the ticket gets issued. A 95% success rate means that out of every 20 tickets issued, 1 will be going to someone who did nothing wrong. Then the burden is on the person to prove they’re innocent, because a machine said they’re guilty.

    • Joe Averkamp August 11, 2016 - Reply

      Edgar, good comment. 95% is a good rate for automation, but not good enough for automated issuance of a ticket. That is why we recommend that after the automated score is developed, the agency conduct a manual review of the images to confirm a violation. In our recommended approach, the manual reviewer would confirm the violation prior to issuing the ticket or adjusting the toll rate. In many toll systems, drivers declaring themselves to be HOV can be charged zero toll,. This system would allow toll agencies to adjust their charges. The intent is to ensure compliance with the rules. While 80-90% of the road users are following the HOV/HOT rules, there are many who are not. The system is intended to restore equity to the roadway—-the vast majority of people follow the rules. Let’s make it fair to them so that everyone is judged by the same rules and people who are bending the rules pay their fair share.

  7. Can this system tell the difference between a person and a seat cover with a picture of a person printed on it?

  8. Joe Averkamp August 12, 2016 - Reply

    The system performs very well in rejecting fake humans as real humans—or in rejecting faces on seat covers.

  9. Why not make them all toll roads along with the automation system. If a single person wants to pay for the privilege let them. If not they get ticketed.

  10. The problem with California Roads & Freeways is CALTRANS they don’t know how to design or build roads . They couldn’t analyze a traffic pattern if it bit them in the butt . I just watched them build an overpass on a main road it took them 18 months to do a 6 month job & they were proud they were 4 days ahead of schedule . I built roads in other states

  11. They turned down red light cameras in Southern California (got rid of them). What makes you like they will incorporate this technology?

  12. Joe Averkamp October 11, 2016 - Reply

    Why Not, I understand the comment. Many people are concerned about red light cameras. When we have spoken with people on HOV/HOT enforcement they seem to view it as an issue of equity. If you are a person driving in the High Occupancy Tolling lanes, and you have paid the toll or have multiple occupants in your car, then you have met your obligation, and the people who are riding in the lane but not paying the toll or riding with multiple occupants are cheating you. This is how people view the HOV enforcement differently. With red lights, everyone at some point in time has misjudged the signal and been caught in the middle of the intersection, and they may have received a ticket. But to drive as a Single Occupant Vehicle in the HOT lane requires planning or lack of planning. The driver may simply have decided that they wanted to take advantage of the rules on the spur of the moment. People whom we have spoken with view it as a matter of equity, if you want to be in the HOV/HOT lane there are steps you can take to qualify. So, people see it as a matter of equity…….pay the toll or ride with others to qualify to be in the lane.

Comments

Your email address will not be published. Required fields are marked *