June 2017

456 78910
18 1920 21222324

Page Summary

Style Credit

Expand Cut Tags

No cut tags
Tuesday, March 21st, 2017 10:21 am
It's getting common to see transportation departments1, 2, 3, 4 and news stories5, 6, 7 advocating zipper merge (AKA late merge), claiming that it is safer and allows higher throughput. As someone who naturally doubts the pronouncements of transportation departments, I thought I'd look into the research behind it.

First of all, what is zipper merge? It's the idea that when lanes merge, the traffic should cluster up at the merge point, where cars will take turns merging. Some variants include active signage to signal which car should go next, but mostly not. This is compared to "early merge", where people start to get over as possible as soon as possible.

Outside of some specific situations (2 lanes going down to 1 where neither lane is obviously the one going away) this has never made much sense to me. Humans overbrake, amplifying any slowdown. Encouraging gradual merging over a long distance, where cars can slow down enough to let others in just by coasting, seems obviously better. And calmer, and more polite. Also, to be fair, the kind of people who are really vocal about zipper merge tend to put me off. But I know better than to trust my gut reaction on such things -- that's what we have science for! So what does the science say?

Under various names, late merge has shown up in papers as early as 199015. The idea really started to take off after some work in the Netherlands in the late 90s.13 Of the later work, two things really stand out to me. First, they almost all are only talking about 2 lanes merging down to 18, 9, 10, 16. Also, many of the papers8, 15, 16, 17, 18 are primarily reporting on simulated results or are lit reviews12, 14 like this post.

The number of lanes in question seems key to me. In the situation of just 2 lanes going down to 1, zipper merge certainly makes a lot more sense. The problem here is that these results are over-generalized to apply to any merging scenario. The entrance to the express lanes on N I-5 in downtown Seattle is a prime example of this. There, a lane becomes exit-only that many people want to take. Many of them apply late merge techniques, zooming up to the front of the line and then trying to merge. In doing so, they block an entire lane that would otherwise be open, and this sometimes carries over to slowing down even the next lane, as people dodge out around the blockage. This is the core of my objection to zipper merge -- it encourages people to be jerks.

The simulation issue is also critical. The simulations are all calibrated using real world data, but that mostly extends just to things like measuring the average time it takes to change lanes. The distinction between real world measurements and simulated results are often lost in later citations, such as in the heavily cited McCoy 200114 paper. It states that "Early Merge has been found to increase travel times", while only referencing two simulation studies15, 16 -- and the second citation only provides parameters for simulating a late merge system that hadn't been implemented yet at the time of publication!

Overall, the results seem pretty muddy to me. Of the others, one9 lacked a control and another10 had neutral results. There are several papers that do find pro-zipper11, 17, 18 results, but those are qualified as being tied to specific traffic volumes and/or the presence of active signage. This is a hard problem to investigate, since it involves changing habits of a large number of people. It shouldn't be surprising if the results are inconclusive.

Does any of this matter? I think so. Look at the news articles promoting zipper merge -- "science says to stop being polite" is a common theme. We should always be extra dubious when evidence seems to justify us doing what we wanted to do in the first place.

1: Minnesota DOT

2: Kansas City DOT

3: Nebraska DOT

4: Missouri DOT

5: Why Last-Second Lane Mergers Are Good for Traffic, New York Times, October 12, 2016

6: All hail the zipper merge: How Canadian politeness is killing the efficiency of our highways, National Post, January 23, 2017

7: Have you ever heard of the zipper merge technique?, Houston Chronicle, July 25, 2016

8: Wakita, Y., et al. "Comparison of zipper and non-zipper merging patterns near merging point of roads." Nature-Inspired Computing Design, Development, and Applications. IGI Global, 2012. 221-231. OPEN ACCESS

9: Grillo, Lia, Tapan Datta, and Catherine Hartner. "Dynamic late lane merge system at freeway construction work zones." Transportation Research Record: Journal of the Transportation Research Board 2055 (2008): 3-10. PAYWALL

10: Idewu, Wakeel, and Brian Wolshon. "Joint merge and its impact on merging speeds in lane reduction areas of construction zone." Transportation Research Record: Journal of the Transportation Research Board 2169 (2010): 31-39. PAYWALL

11: Kurker, Michael, et al. Minimizing User Delay and Crash Potential through Highway Work Zone Planning. No. FHWA/TX-13/0-6704-1. 2014. OPEN ACCESS

12: Walters, Carol H., et al. Understanding road rage: Summary of first-year project activities. No. TX-01/4945-1,. 2000. OPEN ACCESS

13: Dijker, Thomas, and Piet HL Bovy. "Influencing lane changing at lane drops." Transportation Research Board 1999 Annual Meeting CD-ROM. 1999. NOT AVAILABLE ONLINE

14: McCoy, Patrick, and Geza Pesti. "Dynamic late merge-control concept for work zones on rural interstate highways." Transportation Research Record: Journal of the Transportation Research Board 1745 (2001): 20-26. OPEN ACCESS

15: Mousa, Ragab M., Nagui M. Rouphail, and Farhard Azadivar. "Integrating microscopic simulation and optimization: Application to freeway work zone traffic control." Transportation Research Record 1254 (1990). PAYWALL

16: Tarko, Andrzej P., Sreenivasulu R. Kanipakapatnam, and Jason S. Wasson. "Modeling and Optimization of the Indiana Lane Merge Control System on Approaches to Freeway Work Zones, Part I." Joint Transportation Research Program (1998): 345. OPEN ACCESS

17: Ramadan, Ossama E., and Virginia P. Sisiopiku. "Evaluation of merge control strategies at interstate work zones under peak and off-peak traffic conditions." Journal of transportation technologies 6.03 (2016): 118. OPEN ACCESS

18: Kang, Kyeong-Pyo, Gang-Len Chang, and Jawad Paracha. "Dynamic late merge control at highway work zones: evaluations, observations, and suggestions." Transportation Research Record: Journal of the Transportation Research Board 1948 (2006): 86-95. OPEN ACCESS
Tuesday, March 21st, 2017 11:12 pm (UTC)
The 2-to-1 merge is the common case, even in urban areas, so that's the one to optimize. Also, early merge has the potential to push traffic onto surface streets and generally waste pavement. To me, late merge is just common sense. Even if it does encourage jerk behavior - that's psychology, not engineering - you can do early merge without being a jerk, and I think that's how you should.