AI Self-Driving Cars Dealing With Roadway Zipper Merging
About this episode
Lance Eliot breaks down the classic zipper merge debate—whether drivers should “early-emerge” immediately into the open lane or “late-merge” at the last moment to maximize roadway use. He weighs intuitive arguments and notes simulations often favor late merging for throughput, but those studies assume perfectly rational human behavior. The episode then asks whether true self-driving cars could coordinate via V2V (vehicle-to-vehicle messaging) to make merges seamless—while acknowledging the biggest obstacle: humans will still be on the road for years, likely disrupting the AI “bird flock” choreography.
Dr. Eliot explains how AI self-driving cars deal with roadway zipper merging. See his Forbes column for further info: https://www.forbes.com/sites/lanceeliot/
zipper merge
"Okay, let's get started. You undoubtedly know what a zipper merge is, though the name of it might seem unfamiliar. Here's how it goes... This is the classic two-in-one traffic control squeeze play, also known as the zipper merge."
A zipper merge is what you do when one lane ends. Cars from both lanes take turns merging into the open lane, like zipping a jacket closed.
A zipper merge is a traffic pattern used when a lane narrows and drivers from both lanes take turns entering the remaining lane. The goal is to reduce bottlenecks and keep traffic moving smoothly rather than everyone trying to force themselves in early or late.
lane narrowing
"Imagine you're driving along on the highway, minding your own business, when you notice up ahead an indication that your lane is being narrowed out and you'll need to get over into the other lane next to you."
Lane narrowing is when the road gets smaller—like when a lane ends because of construction. The car has to notice that the lane is ending and merge safely.
Lane narrowing is the roadway condition that triggers zipper merging, typically caused by construction zones or temporary barriers. For self-driving cars, it’s a key perception and planning trigger because the vehicle must recognize the end of a lane and execute a safe merge.
electronic board side might be flashing warnings
"...up ahead an indication that your lane is being narrowed out and you'll need to get over into the other lane next to you. Electronic board side might be flashing warnings that your existing lane is going to end soon."
Those flashing signs are electronic boards that tell drivers what’s happening ahead, like a lane ending. A self-driving car needs to understand those warnings too.
Electronic message boards are variable traffic control devices that provide real-time instructions or warnings to drivers. In the context of self-driving cars, these signals can be part of the information the system uses to anticipate lane closures and required driver actions.
red cones
"In addition, a series of weather-worn red cones are set up in the lane to inch you over step by step."
Cones are used to guide traffic through a work zone. They help show where lanes are and where cars should go when merging.
Traffic cones are temporary channelizing devices used to guide vehicles through construction zones and lane closures. For automated driving, cones are important visual cues for lane boundaries, drivable space, and the intended merge corridor.
two-in-one traffic control squeeze play
"In addition, a series of weather-worn red cones are set up in the lane to inch you over step by step. This is the classic two-in-one traffic control squeeze play, also known as the zipper merge."
It’s a highway situation where two lanes get forced into one. Drivers have to merge carefully because there’s less room than usual.
This describes the common highway situation where two lanes become one due to construction, lane closures, or road geometry. The “squeeze play” framing highlights how drivers must coordinate to merge efficiently under constrained space.
heated debates
"So which camp do you fall into? The early-merge member or the late-merge member of our world? That's where heated debates start to occur."
People argue about whether you should merge early or wait until the last moment. A self-driving car has to pick a strategy that works with what other drivers expect.
The “heated debates” refers to the real-world disagreement between early-merging and late-merging behaviors. For autonomous driving, this matters because the system must choose a policy that is safe and socially compatible with how human drivers behave.
late-merge
"In contrast, the late-merge types are likely to contend that the early-merge drivers are skittish and too timid... The late-merge logically allows for better use of the roadway availability."
Late-merge means waiting until later in the merge zone to switch lanes. The benefit is that both lanes get used longer, but it only works well if drivers take turns instead of forcing their way in.
Late-merge refers to waiting until closer to the merge point before entering the single lane. The argument for late-merge is that it can improve throughput by keeping both lanes available longer, but it requires drivers to cooperate to avoid sudden braking and conflict.
early-merge
"Indeed, when the early-merge purist meets with the late-merge perfectionist... The early-merge logically allows you for a more measured approach that calmly allows for drivers to exit from their existing lane and comfortably merge to the next lane over."
Early-merge means getting over before the merge area, instead of waiting until the last moment. People argue it can feel calmer, but it can also cause congestion if too many cars do it early.
Early-merge refers to drivers leaving their current lane before the merge point to get into the other lane sooner. In merge debates, early-merge is often argued to reduce uncertainty and allow smoother gaps, but it can also reduce lane utilization if done too aggressively.
2-in-1 traffic condition
"Indeed, when the early-merge purist meets with the late-merge perfectionist during a 2-in-1 traffic condition, there's a sizable potential for sparking road rage."
A “2-in-1” situation is when two lanes come together and become one lane. That’s where drivers often get impatient because everyone has to squeeze into the same space.
A “2-in-1” condition describes a roadway bottleneck where two lanes combine into one. These merges create conflict points and can amplify aggressive behavior if drivers don’t coordinate their timing.
traffic strategy
"The debate about which driving strategy is right and which is wrong has been going on since the invention of the zipper merge."
A traffic strategy is just the way drivers choose to handle a tricky situation like merging. Here, the “strategy” is whether you move over early or wait until later.
A traffic strategy is a driver behavior pattern used to navigate flow constraints like lane merges. In this segment, the strategy is framed as either early-merge or late-merge, with different assumptions about how drivers create gaps and how that affects safety and efficiency.
roadway availability
"The late-merge logically allows for better use of the roadway availability. If drivers prematurely get out of an available lane, they're going to crowd into the other remaining lane..."
Roadway availability is basically how much “space and time” is still available for cars to move. If too many cars leave one lane too soon, the other lane gets crowded faster.
Roadway availability refers to how much usable capacity remains in each lane before the merge point. The late-merge argument claims that leaving a lane early reduces the effective capacity of the remaining lane, which can increase overall congestion.
simulations
"Numerous simulations have been undertaken to try to figure out the proper choice. By and large, the simulations tend to suggest that the waiting to merge is the better option..."
Simulations are computer experiments that try to predict what traffic will do. The point here is that the computer results may not match real life because real drivers don’t always behave perfectly.
Simulations are computer models used to test traffic behaviors under controlled assumptions. Here, the speaker notes that many simulations favor waiting to merge, but they often don’t capture real human driving variability.
AI-based true self-driving cars
"Now, the future might provide a kind of resolution. Consider this intriguing point. Will the advent of AI-based true self-driving cars make zipper merges into an easy-peasy situation..."
This refers to cars that can drive themselves using AI. The idea is that they might merge more smoothly and predictably than people, which could make zipper merges less stressful.
AI-based true self-driving cars are fully automated vehicles that use artificial intelligence to perceive the environment and control driving actions without human intervention. The speaker is suggesting that such systems could coordinate merges more consistently than humans, potentially reducing conflict in zipper-merge scenarios.
level four and level five true self-driving cars
"For level four and level five true self-driving cars, there won't be a human driver involved in the driving task. All the occupants will be passengers."
These are categories for how “self-driving” a car is. At Level 4 and Level 5, the car is designed to handle the driving itself, not rely on a person to steer or drive.
“Level 4” and “Level 5” are SAE autonomy levels describing how much of the driving task the automation system handles. At these levels, the system can perform the driving without expecting a human to take over for the driving task (Level 4 in defined conditions; Level 5 universally).
V2V, vehicle-to-vehicle electronic messaging
"This is accomplished via the use of V2V, vehicle-to-vehicle electronic messaging."
V2V means cars can “talk” to each other wirelessly. That helps them coordinate moves like merging so it feels more orderly and less chaotic.
V2V (vehicle-to-vehicle) communication lets cars exchange data directly with nearby vehicles. For maneuvers like a zipper merge, V2V can share intent (who is merging, when, and how), enabling smoother coordination than relying only on sensors and human behavior.
AI-based self-driving car could coordinate with each other
"Each AI-based self-driving car could coordinate with each other during a zipper merge and really handle the merging activity by communicating to each other via the V2V..."
Instead of each car merging “on its own,” the cars work together. They can line up their timing so the merge happens smoothly.
The key idea is multi-agent coordination: multiple autonomous vehicles coordinate their timing and lane changes to create a smooth merge. With V2V, cars can coordinate intent and spacing, reducing gaps and conflicts that typically cause zipper-merge slowdowns.
human interlopers will mess up this driving dance
"They will also need to contend with human drivers. And you can pretty much bet human interlopers will mess up this driving dance."
If some cars don’t merge in the expected way, it can throw off the whole plan. Self-driving cars have to be able to handle those “surprise” moves.
This is about interaction effects in traffic: autonomous coordination can be disrupted by vehicles that don’t follow the expected merging protocol. In practice, that means self-driving systems need robust planning for edge cases—like late lane changes, aggressive gaps, or inconsistent yielding.
Request an Explanation
Heard something you'd like explained? We'll add it to this episode.
Sign in to request explanations for terms you heard.
Want to learn more?
Browse our glossary for plain-English explanations of automotive terms, jargon, and concepts.
Help improve this episode
See something that's not quite right? Our annotations are AI-generated and can sometimes miss the mark. Click the flag icon on any annotation to suggest a correction.