Traffic is much more than just a swarm of wheels on asphalt moving in a somewhat coordinated way. It’s actually one of the most protocolized environments that most ordinary people enter every day. It’s a system of rules, signals, roles, expectations, and fallback behaviors that allow many independent actors to coordinate without constantly renegotiating with each other.
Road networks aspire to have everyone using them roughly agree on who has priority, what different signals mean, where each actor belongs, what behavior is expected, and what rules are suspended in case a traffic light stops working.
So, it’s safe to say that drivers don’t simply operate vehicles. They also anticipate other people’s behavior, their hesitation, impatience, courtesy, aggression, and confusion. Even an occasional wave through the windshield. Traffic systems are based on protocols, rules, and implied communication between different participants.
But what happens when you add a participant with its own set of operating protocols and an inability to communicate with other drivers and interpret unusual traffic patterns?
On November 8, 2017, Las Vegas launched a public autonomous shuttle in its downtown innovation district. The shuttle, integrated with smart-city traffic infrastructure, ran a short loop through downtown streets, successfully communicating with traffic signals, while carrying passengers in ordinary public traffic.
However, just hours after launch, the shuttle encountered a truck backing into an alley on South 6th Street. It detected the truck and came to a stop but it was struck by the truck, which hadn’t stopped reversing.
This is the simplest version of the story, and while the autonomous vehicle ended up with just a minor scratch on its fender, the point is that the shuttle did what it was supposed to. The human driver didn’t. That version of the story, while technically accurate, also doesn’t capture the whole picture. A better explanation would be that two different rule systems met on the same street, and they failed to coordinate.
The Street Before the Shuttle
As previously stated, street traffic is one of the most protocolized environments most people interact with, and South 6th Street was no exception prior to the incident. It had its own speed limits, lanes, parking spaces, intersections, delivery access, alleys, and traffic signals. These are all the things that shape human behavior in traffic, basically forcing drivers to operate inside a dense structure of rules and expectations.
However, the streets also depend on informal coordination and communication. A delivery driver backing into an alley isn’t only following traffic rules. They’re also checking mirrors, watching pedestrians, judging clearances, and assuming that surrounding drivers will respond to these actions in recognizable ways.
A human driver who sees a truck backing up into an alley may slow down earlier, stop further back, honk, reverse, wait, or communicate with the driver, either verbally or through gesture. Some of those actions derive from rules, but others are social habits developed around those rules. There are some very important distinctions between the two.
Human drivers follow rules, but also interpret different situations, and adjust their actions accordingly. Some rules, like right-of-way, can be bent or waived; a human driver can yield their right-of-way in certain situations to avoid creating a jam, thus ensuring smoother traffic flow. A perfect example of this would be stopping early at a gridlocked intersection to allow another driver to turn out of a blocked driveway or side street.
A human driver may adhere to the rule or choose to bend it out of courtesy towards another driver, or in an attempt to restore order to the system. But a machine, such as an autonomous shuttle, enters the traffic with a different kind of discipline. The machine doesn’t improvise in the human sense. It follows its designed responses to detected conditions and executes bounded behavior.
This is the advantage of autonomous vehicles, as they don’t get distracted, impatient, tired, offended, or overconfident. But the downside is that they can’t actually automatically participate in the full social exchange of street traffic.
The Las Vegas shuttle wasn’t just an autonomous vehicle, but a moving protocol inserted into a public environment dominated by human norms. And humans, compared to machines, are much more prone to errors.
What Did the Shuttle Do?
The collision wasn’t caused by the shuttle failing to see the truck. According to accident records, the shuttle’s sensor system detected the truck from a distance and tracked it as it reversed. Then, it stopped three meters (approx. 10 feet) from the obstacle in its path. As the truck backed into the alley, the shuttle simply decelerated and came to a stop.
So, the shuttle did nothing wrong; it executed its programming flawlessly. It detected an obstacle, it reduced its speed, and eventually stopped. The collision risk was mitigated as far as the shuttle’s programming and rules allowed. From the machine’s perspective, this was a success.
But from the perspective of street traffic, the event wasn’t over. The truck was still moving, and the human driver inside the truck was still backing up. The shuttle’s passengers and attendant could clearly see the danger unfolding as the driver continued backing the truck into the autonomous vehicle. According to some reports, they even waved toward the truck driver.
However, the truck driver continued reversing, striking the shuttle several seconds later. This is where the protocol problem appears. The shuttle clearly followed programming that made it avoid moving into the truck. But the shuttle’s internal programming didn’t know how to make the truck stop. It couldn’t negotiate with the driver of the truck in a fully human way.
It couldn’t lean on the horn as most humans would; it didn’t reverse out of the way like human drivers might’ve tried to do, or make itself known as a participant in traffic. Instead, the shuttle was practically an object that had simply stopped and positioned itself in the truck’s way.
Regardless, the shuttle obeyed its safety logic, but the street traffic required a different kind of coordination. Perhaps even non-verbal communication, like leaning on the horn.
A Machine Rule Meets Human Assumption
The truck driver, on the other hand, wasn’t navigating an abstract system that he hadn’t encountered before. He was doing a familiar delivery maneuver on a route he knew well. Furthermore, he had experience with the area and with the practical difficulty of moving a long vehicle into an alley.
According to the investigation, he was paying attention to parked vehicles and a pedestrian while backing. He also saw the shuttle and assumed it would stop at a reasonable distance from the truck. And that assumption is the very heart of the case.
In traffic, a “reasonable distance” isn’t purely mathematical. It’s also partially social. A driver expects another driver to understand the geometry of the situation and the purpose of the maneuver the other driver is performing. A backing truck isn’t just an obstacle, but a large vehicle performing a slow, often awkward, and partially obstructive act that other drivers often have to accommodate.
Sometimes, accommodating means backing up a bit and creating enough space for the truck driver to complete the maneuver. The shuttle, on the other hand, did accommodate the maneuver, but only according to its formal rule set by stopping at the programmed distance. The truck driver might’ve expected a kind of spatial cooperation that was wider, more flexible, and more intuitive than the shuttle’s stopping protocol.
None of the involved parties behaved recklessly; each system just modelled the situation differently. That in itself is the lesson here: protocols shouldn’t be correct only when operating in isolation. They should be mutually legible and interoperable.
Michael P. McCready, founder of McCready Law, frames the crash in similar terms, as a failure of coordination rather than a simple failure of automation. “The shuttle did the most machine-like thing possible: it followed its protocol,” McCready says. “The truck driver did the most human thing possible: he operated in a world of habit, assumption, and incomplete information. The collision happened in the gap between those two systems.”
That gap, McCready argues, becomes especially important when autonomous vehicles enter ordinary public spaces. “As autonomous vehicles move into public spaces, safety is not just a software issue; it is a coordination issue. The technology has to be sound, but so do the rules, infrastructure, communication signals, and human responsibilities around it.”
The Limits of Stopping
Stopping is one of the most basic safety behaviors in both automation and traffic. If the system detects an obstacle, it stops, which makes sense, especially for low-speed public shuttles carrying passengers. A cautious stop is predictable, conservative, safe, and easily defended.
But stopping doesn’t necessarily resolve a traffic conflict, and human drivers understand that intuitively. While stopping often prevents danger, stopping at the wrong place at the wrong time can also create it. Sometimes, the safest course of action is to actually reverse, move aside, signal more aggressively, or create space for another road user to complete their maneuver.
These decisions are messy because they often involve prediction, social reading, and judgment about what others are likely to do next. Human drivers also build rough profiles of other road users, consciously or otherwise, based on the vehicle they drive, their speed, hesitation, positioning, and other behavioral cues. A learner, an aggressive driver, or someone who appears uncertain may cause us to adjust our own behavior before any formal traffic rule requires us to do so.
In that sense, driving depends on an interaction between protocol and profile. The Las Vegas shuttle protocol was designed around avoiding active movement into an obstacle, and, as previously mentioned, that’s really important. However, the vehicle had to deal with an object that was moving towards it, which isn’t the same thing.
Smart Infrastructure, Ordinary Streets
The Las Vegas shuttle crash is a significant event, because it shows that a smart infrastructure, the one in which an autonomous vehicle has to communicate with traffic signals, doesn’t make the whole street smart. A connected traffic signal can tell a shuttle how long a light will remain green, but it can’t guarantee that a truck driver backing into an alley understands how the shuttle might behave.
It can’t make human actions part of the same communication network, thus revealing the real limits and boundaries of protocolized infrastructure. Formal systems coordinate well with other formal systems, but the streets also contain actors who communicate through habit, not data packets.
The truck wasn’t part of the shuttle’s digital conversation with the city and the smart infrastructure, not part of a protocol layer, but a part of a physical street. And that difference is easy to miss when smart-city systems are described as if connectivity itself produces coordination. It doesn’t; connectivity helps, but only when relevant actors are connected and able to understand each other.
Responsibility in a Mixed System
Now for a million-dollar question: who is to blame for the crash? In this case, the responsibility was relatively easy to assign, and the truck driver was cited. There were no injuries, and the damage was quite minor.
But protocol responsibility is less simple than finding who’s to blame. If the autonomous vehicle stops and still gets hit, what should change to prevent that from happening again? Should we change the software or the road design? Traffic rules? The assumptions humans make about autonomous vehicles?
That last question may become increasingly important. Human drivers already recognize rough profiles on the road, such as the hesitant learner or the aggressive driver, and adjust their expectations accordingly. As autonomous vehicles become more common, “autonomous vehicle” may become another such profile, with human drivers learning to expect particular kinds of caution, predictability, and limited improvisation from them.
The answer may be some version of all of the above, because autonomous systems tend to redistribute responsibility. A human driver is still responsible for backing safely, and a shuttle operator is still responsible for safe deployment, and the city is responsible for the public environment in which traffic systems operate. The Las Vegas shuttle incident shows that those responsibilities overlap.
Final Thoughts
The crash attracted plenty of attention due to its timing. The autonomous shuttle was new, public, and meant to demonstrate the future of urban mobility. Sure, the damage it suffered was minor, but the symbolism of such an event was much larger, and it became evidence in a broader argument about whether the autonomous vehicle technology was ready.
Such treatment is unfair because people forget that human drivers are also part of an infrastructure, because the machine followed protocol, and the human made the mistake. However, things aren’t so black-and-white, and while humans are more prone to errors, they’re also part of the traffic system autonomous vehicles need to learn to recognize.
Despite the fact that dramatic failures invite dramatic conclusions, the lesson here isn’t that autonomous vehicles aren’t safe or that human drivers are hopeless. It’s that protocolized environments only work when their users understand what the protocol expects from them.




