Subscribe to DSC Newsletter

Optimizing highway traffic, particularly during peak times, is a high priority for most local governments. Many policies have been tried to improve the situation, for instance, HOV lanes, adding more lanes, switching lanes morning/afternoon, public transportation, tolls and fees, and telecommuting. Most of these solutions have proved to be a failure. 

Trucks represent 5% of the traffic at commute times (in terms of number of vehicles), and maybe 20% of the highway space occupied by idle or moving vehicles. Because they take a very long time to start moving and picking up some speed (or conversely, when they brake), and because morning or afternoon commute times is just stop and go on the highways in all suburban areas, I believe trucks are responsible for increasing everyone's commute time by more than 100%, despite the fact that they represent only 5% of all vehicles at commute times.

A possible solution is to reduce the number of trucks (especially the big ones) from current levels to less than 1%, at commute times. How to achieve this goal? Proposed solution: Have truck driver employers provide an incentive, such as

  • Much better pay for reducing miles/gallon for truck drivers (the morning/afternoon commute times must cost them tons of money)
  • Traffic alerts and data science products that tell truck drivers when and where to drive to optimize their routes
  • Trucks on auto-pilot (automated driving)?

What are your suggestions? How do you asses a dollar amount and responsibility share for problems created by big trucks slowing down morning traffic to a crawl? Is this a real problem, or is the real cause very different (maybe use Monte Carlo simulations to measure impact of trucks on commuting traffic)? What other solutions do you envision? How can data science help? How about educating all car drivers to have faster reflexes (taking 0.1 second to start after a stop, rather than 1.0 second) or maybe get an automated auto-start feature installed on all cars? How about further optimizing / synchronizing traffic lights?

Related articles

Views: 3463

Reply to This

Replies to This Discussion

Another thought: bidding system for trucks' operating hours, though industries will almost definitely lobby against such an implementation. Also, would we consider many other types of indirect optimisations such as resource sharing to reduce the absolute number of trucks on the road at any one time? If truck services can be outsourced in an efficient and reliable manner, companies will save physical assets maintenance cost.
An interesting speculation. Do you have data and simulations to support this?

I used to race bicycles, and there, where all vehicles have near identical performance characteristics, we would observe the "concertina" effect on all corners. This is caused by the front riders slowing, then accelerating through corners. The following riders also slow, and each one has further to accelerate (reaction times and allowing safety distance margins).

My point is that the effect has been observed, and simulated (at Los Alamos Labs) independent of there being trucks.

Data is what is needed to support speculations.

@Alex: Nope, it is just a wild guess and I could be proven all wrong. I'm wondering if someone would be interested in checking my assumption using Monte Carlo simulations, and see if the gain is as big as I think (when you eliminate trucks) or negligible. 

I respectfully disagree.  Based on years of observation, trucks improve average speed. Auto drivers cause reduced average speed.  Auto drivers drive at random and frequently variable speeds all optimizing different programs.  Truck driver is trying to stay on a schedule, consistant speed, predictable location for several hours.  Any improvement in average speed will increase wait time at the destination.  Being late may mean a missed loading dock slot, delay, job loss.  Speeding is not an option.  His schedule has already factored in the average traffic delay caused by automobile drivers not paying attention, making poor decisions, causing collisions, etc.  It is random variation that is the problem, not trucks.  Think M/D/1 model vs M/M/1.  D/D/1 would be even better.

I drive the same stretch of highway each time I go to the office.  I keep right, go the speed limit whenever I am able, leave 1 car length / 10 MPH of speed, let people in left or right.  I enjoy almost every day, seeing a flashy car blow by me and seeing it again stopped on my left before I make my exit.

I'm going to do some simulations or math computations to see what the gain could be. Another idea (if trucks are < 5% of commute traffic) is force them to use HOV lanes during commute time. And maybe with one entry/exit to HOV lane every two miles, or every mile, no more than that (they are sometimes called express lanes).

A reasonable mathematical model could work as follows:

1)  Two lanes, trucks and car in each lane

2) Passing is done only to change lane to exit highway, not to gain speed

3) Vehicles moving at same speed on both lanes, on average

Under this model, the speed is determined by that of the slowest vehicles: the big trucks. Assuming they take 3 times as much time than cars to move from 0 to 40 miles/hour, and assume the 40 miles/hour is the upper bound because of traffic overload, and that you get stopped every 0.5  mile, I would imagine trucks slow traffic by a factor two, especially on highways with 10 or more miles of traffic jam.

I know of large cities that restrict entry of trucks during the peak hours. But this gives rise to a new queuing challenge since the trucks have to be parked outside city limits till the off-peak hours. The best way of reducing traffic congestion is having a good rail system for mass transit.

Excellent point. Real world experience indicates that truck drivers are better able at modulating speed and minimize the "slinky effect" while most drivers will tap their brakes causing a domino reaction. So I always to choose to follow 18 wheelers for LESS stop-n-go, not more ( I drive a manual and the clutch is tedious in traffic). We just need to eliminate brake lights to force people to observe changes in speed instead a Pavlovian reaction to lights.

I think the quick changes in speed is what leads to unnecessary traffic, not the gradual ones from trucks. So I offer that the cars are more the culprit.

Trucks have a larger mass and have lower acceleration to reach the same velocity as cars as Vincent mentioned. (Same power being derived per litre of gasoline to provide a higher momentum etc.). This would contribute to the longer time in the queue. A single truck causes more delay in the queue than a single car. However, if we started replacing the trucks with the cars, we’d have to account for the “reaction time” of the additional drivers which causes the accordion / slinky effect. I suspect there will be a break-even point where have x number of cars is just as bad as having a  truck sharing the road in a stop-start situation.

I am interested if anyone has modeled the slinky effect of traffic flow after modern suspension. Think about the compression and rebound dampening circuits of a shock absorber. The purpose of the oil flow through valves is to dampen or control the motion so that the suspended vehicle remains stable and the motion is smooth. That is also the purpose of metering lights at on-ramps. So I again offer that increased acceleration and deceleration (more due to the drivers than the type of vehicle) worsens traffic. Stay off the brakes and cruising in first gear will even out the flow. Now it bears to make a distinction between traffic jams and volume. The latter is simply due to the number of vehicles and the amount of space. If you've ever witnessed the mass start of the Hawaii Ironman with 1700+ swimmers vs. other races, you will understand that better self-seeding leads to smoother flow and faster times.

My son and I used to discuss topics similar to this when he drove to high school. Now that he has a Bachelors Degree in Computer Science, the topic still comes up. I sent him the first link, he spotted the second, and sent me the third from another source.

From the sidebar:

The last page of the trafficwaves link has links to some simulations.



  • Add Videos
  • View All

© 2019   Data Science Central ®   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service