Recalculating! By not driving the optimal route, you're causing traffic jams

March 16, 2016 by Marta González And Antonio Lima, The Conversation
Drivers make some suboptimal routing decisions when they’re traveling around town. Credit: A. Lima et al. J. R. Soc. Int. DOI: 10.1098/rsif.2016.0021, Author provided

If you use a car to get around, every time you get behind the wheel you're confronted with a choice: how will you navigate to your destination? Whether it's a trip you take every day, such as from home to work, or to someplace you haven't been before, you need to decide on a route.

Transportation research has traditionally assumed that drivers are very rational and choose the optimal route that minimizes travel time. Traffic prediction models are based on this seemingly reasonable assumption. Planners use these models in their efforts to keep traffic flowing freely – when they evaluate a change to a road network, for instance, or the impact of a new carpool lane. In order for traffic models to be reliable, they must do a good job reproducing user behavior. But there's little empirical support for the assumption at their core – that drivers will pick the optimal route.

For that reason, we decided to investigate how people make these choices in their real lives. Understanding how drivers build a route to reach their destination will help us gain insights into human movement behavior. Better knowledge of individual routing can help improve urban infrastructure and GPS directions systems – not just for one driver, but for everyone. Beating congestion is a big goal: one estimate put the cost of traffic in 2014 at US$160 billion in the U.S., with 42 extra hours of travel time and $960 worth of extra fuel for every commuter.

A sample of the transformed trajectories reveals the shape of human routes. Regardless of the real start and destination points, every transformed trajectory begins at the circle on the left and ends at the circle on the right. Credit: A. Lima et al. J. R. Soc. Int. DOI: 10.1098/rsif.2016.0021, Author provided

How do people really go?

Using GPS data collected for several months for hundreds of drivers in four European cities, we studied individuals' routing behavior, looking for interesting patterns in their choices.

We discovered that people use only a few routes when moving between their relevant places, even when those trips are repeated again and again over extended periods. Most people have a single favorite route for trips they perform routinely and a few alternatives routes they take less frequently to the same destinations.

So did people in fact usually choose the optimal route?

In short, no. It turned out roughly half of the favorite routes are not the suggested by navigation devices, such as those offered by some popular mapping apps for smartphones. If we also consider drivers' alternative choices, even fewer routes are optimal – only a third overall minimize travel time.

Our data provide empirical proof that drivers are not taking the optimal route, directly contradicting the shortest travel-time assumption.

The density plot shows how likely you are to be at any position between the start (on the left) and the destination (on the right). Colors indicate, in logarithmic scale, from dark to bright, the spots more likely to be occupied by drivers on that trip. Credit: Antonio Lima, CC BY-ND

Why would drivers take a nonoptimal route?

What's behind this result? A unique answer that is valid for every driver won't be easy to find.

Prior small-scale studies found that many factors, some seemingly minor, might influence route preference. For example, people tend to choose routes going south rather than routes of equal lengths that go north. People favor routes that are straight at the beginning, instead of shorter ones that aren't straight. Landmarks also influence route choice, by attracting more trips than minimization would expect. A novel app for iPhones builds on that very concept and allows people to find the most "interesting" route between two points.

People might not be able to determine which route is optimal, among all possible choices, because of limited information and limited ability to process big amounts of information. Or, even if they can, people might deliberately make different choices, according to personal preference. Many factors can influence preference, including fuel consumption, route reliability, simplicity and pleasure.

Drivers' apparent flexibility on route choices may provide an opportunity to alleviate overall congestion. For instance, smartphone apps could offer points and vouchers to drivers who are willing to take longer routes that avoid congested areas. Navigation app Waze has already changed drivers' habits in some cities, so it's not so far-fetched to imagine a gamification system that reduces congestion.

How far from the best route are we?

For our next study, rather than trying to understand what drives individual route choices, we aimed to quantify how far those choices are from optimal.

It's hard to directly compare all the different trips undertaken in a city, because they involve many locations and are different in length. To make this task easier, we transformed trajectories so that they all look alike, regardless of their actual source, destination and length. We rotated, translated and scaled each route so that all trajectories would start and finish at the same two points in a new reference system. After this transformation, all the routes look as if they spanned the same two points; they all look similar in length, but their shape is preserved. What we found by plotting a sample of the transformed routes was the intrinsic variability in human routes.

Intriguingly, our abstraction of all the trips sort of looks like a magnet's force lines, with the routes' origins and destinations in place of the magnet's north and south poles. By analyzing a density plot of the transformed trajectories, we found the vast majority are fully contained within an ellipse that has the same shape independent of the scale, with the start and endpoints as foci. This ellipse effectively makes up the boundary of human routes.

The ellipse also helps us measure how direct a route is. The ellipse's eccentricity tells us how elongated it is. An eccentricity close to 1 means the ellipse is similar to a line (high width and low height), while an eccentricity close to 0 means it is similar to a circle (width and height roughly similar).

Generally, a straight is not a viable option because of physical obstacles, such as buildings. Drivers deviate from that idealized shortest path according to the street network and personal preferences. While these two phenomena are hard to model, we found that they are bounded by a ellipse of a particular shape, having a high eccentricity equal to 0.8.

To our surprise, the observed shape of the ellipse did not change with distance between the endpoints. It looks like in an urban setting, drivers are willing to take detours that are roughly proportional to the distance between their starting point and destination. Routes that involve bigger detours are simply not taken, or split into two separate trips.

Our study uncovered basic rules of a realistic routing model that captures individual behavior in a urban environment. These findings can be used as building blocks for new routing models that better predict traffic. And now that we know have some quantifiable flexibility in their routes, we can use this information to design incentive mechanisms to alleviate congestion on busier roads, or carpooling plans based on individuals' preferred routes.

Explore further: Uber looks to get commuters out of their personal cars with new fixed route option

More information: Antonio Lima et al. Understanding individual routing behaviour, Journal of The Royal Society Interface (2016). DOI: 10.1098/rsif.2016.0021

Related Stories

How wind might impact birds' migration routes

October 19, 2015

For centuries, scientists have been working to unravel the many mysteries of bird migration, studying where birds go, how they find their way, and how much of the information they need is inherited and how much is learned.

Cutting electric vehicle energy use 51 percent

September 9, 2014

( —Researchers at the University of California, Riverside's Bourns College of Engineering have shown that a vehicle navigation tool they created can cut electric vehicle energy use up to 51 percent.

Smart satnav drives around the blue highway blues

February 7, 2013

Endlessly frustrated by congested roads, computer scientists at California State University, in Fullerton have developed a satellite navigation system, GeoTNavi, which hooks into historical traffic data and current vehicle ...

Recommended for you

What can snakes teach us about engineering friction?

May 21, 2018

If you want to know how to make a sneaker with better traction, just ask a snake. That's the theory driving the research of Hisham Abdel-Aal, Ph.D., an associate teaching professor from Drexel University's College of Engineering ...

Flexible, highly efficient multimodal energy harvesting

May 21, 2018

A 10-fold increase in the ability to harvest mechanical and thermal energy over standard piezoelectric composites may be possible using a piezoelectric ceramic foam supported by a flexible polymer support, according to Penn ...

Self-assembling 3-D battery would charge in seconds

May 17, 2018

The world is a big place, but it's gotten smaller with the advent of technologies that put people from across the globe in the palm of one's hand. And as the world has shrunk, it has also demanded that things happen ever ...


Adjust slider to filter visible comments by rank

Display comments: newest first

5 / 5 (4) Mar 16, 2016
Why would drivers take a nonoptimal route?

Here's a few additional factors:

1) Stop-and-go vs. continuous traffic. Example: On my way to work I can take two routes until I reach the Autobahn. One has eleven traffic lights the other has three. Even though the one with eleven traffic lights is shorter it requires a lot more attention. Especially during rush hours when people don't drive at their best (due to being tired, stressed, late, whatever).

2) (which plays directly into 1) ) : Potential for accidents. People prefer routes that do not include accident prone locations (e.g. left turns at busy intersections, narrow streets with oncoming traffic, etc.) instead may favor a route that is more circuitous but safer.

3) People prefer routes that have construction less often. Even if there is no construction currently going on people don't like to find out when it's too late - so they may avoid frequent construction routes per default
Multivac jr_
not rated yet Mar 16, 2016
It's advantageous to know multiple ways to get from one point to another since there will be times when our primary/most-efficient route will be unusable.

And by "knowing" various routes I mean by memory, not simply by being able to follow alternate GPS directions.

Also, some people like to avoid speed cameras/traps.
2.5 / 5 (2) Mar 16, 2016
Another one is people tend to stay away from more direct routes when the route takes them through the worst section of town. 5 minutes extra driving is a lot better than getting robbed or killed just because you happen to drive though some dirtbag gang's turf.

not rated yet Mar 17, 2016
What does "optimal" routing mean? Is it the shortest distance? The safest route? The least amount of time? Easiest to navigate?

The authors don't actually define the term all that well. Then they go out on a limb and say that people don't drive the routes they expect. Then they get really stupid and start thinking that perhaps they can encourage people to use whatever algorithm for "optimal" that they may have in their pointy little heads.

Perhaps they should try asking drivers why they chose a particular route instead of trying to "optimize" public behavior to their way of thinking. They might actually get to whatever goal they're seeking by just talking to people.
5 / 5 (1) Mar 17, 2016
What does "optimal" routing mean? Is it the shortest distance? The safest route? The least amount of time? Easiest to navigate?

The authors don't actually define the term all that well.

I'd infer that they mean the shortest route from this part:
" People favor routes that are straight at the beginning, instead of shorter ones that aren't straight."
Thirteenth Doctor
5 / 5 (1) Mar 17, 2016
In Florida, people do not know how to merge at all and this makes many routes unfeasible on my commute to work. I'd rather take the extra time without people who aren't responsible enough to make ensure they have enough time to get to work so that they don't have to drive as dangerously.

Not to mention the many rear ends I've seen happen because some one is trying to speed around three cars and quickly change lanes only to have to suddenly brake.
not rated yet Mar 17, 2016
I'd infer that they mean the shortest route from this part:
"People favor routes that are straight at the beginning, instead of shorter ones that aren't straight."

Except that shorter routes don't necessarily get you to a destination faster. There may be more traffic lights, the speed limits may be lower. Basically, your overall fuel expenditure can easily be worse on a shorter route. I have a situation like that in my commute.
not rated yet Mar 20, 2016
All too often the "best" route for shorter distances includes cuts through residential neighborhoods or a ridiculous number of turns. For longer distances, algorithms attempt to minimize time and you can end up driving much longer distances to use highways. An example of the first: coming home from a certain restaurant my gps wants me to make five turns before I get to the main road, I can get to that road by staying on the first street for a few extra minutes and making ONE turn, and that a traffic light.
One thing I learned very early in my gps use: NEVER select shortest route!
not rated yet Mar 20, 2016
The optimal route is no longer the optimal if everyone tried to take it, because it would become congested.
1 / 5 (1) Mar 21, 2016
Most of the time, the shortest routes take the longest time as well as greater use of fuel due to stop & go features of the shortest routes. The density model was more like an electron cloud feature of an atom.

No one with an ounce of common sense would embark on a trip unconsciously engaging in driving behavior the density model predicts for human behavior in traveling from/to the featured points.

The model should like more like a very flat ellipse existing only between the two designated points instead of the fat oval ellipse extending far beyond the boundaries of the two designated points. Nothing like embellishing a model.
not rated yet Mar 21, 2016
Hmmm.... so the road less traveled is optimal.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.