Passenger Trains > Passenger flees wayward Waymo on Phoenix tracks

https://www.azfamily.com/2026/01/08/waymo-passenger-flees-after-car-drives-phoenix-light-rail-tracks/ Waymo passenger flees after car drives on Phoenix light rail tracks

------------------------------------------------------------------- 

Mickaela Castillo1/8/26. *PHOENIX (AZFamily)* —

A Waymo was caught on camera driving down Phoenix light rail tracks this week, forcing a passenger to flee the car before it continued along the tracks near an oncoming train. Video taken by a bystander near Central and Southern avenues shows the moment the self-driving car stops on the tracks just before an oncoming light rail approaches. The passenger runs out of the vehicle before the car continues to drive down the tracks near another train. 

“I actually felt a little sorry for the car. It obviously made a bad decision and got itself in a difficult place,” said Andrew Maynard, an emerging and transformative technology professor at ASU ( https://search.asu.edu/profile/2670673 ). Maynard said while these situations are rare, they do happen. “This is exactly one of those edge cases, what we call them. Something unexpected where the machine drove like a machine rather than a person,” Maynard said. 

Waymo vehicles are equipped with 29 cameras and update their routes and systems weekly. However, in the area where the incident happened, there was construction and the light rail was added to this spot within the last year, which Maynard said could have contributed to the rail track detour. 

“I think Waymo has a challenge because no matter what they do with their system, there are always going to be unexpected circumstances where they have to learn from them,” Maynard said. 

While incidents like this may draw skepticism about how safe these cars are, Maynard said he believes they’re likely safer than typical human drivers. “These cars, in many circumstances, are safer than human drivers because they don’t have distractions, like a human driver does,” Maynard said. 

Valley Metro released the following statement:  > “At approximately 9 a.m. on Wednesday morning, a Valley Metro employee> observed an autonomous Waymo vehicle on the northbound light rail tracks> near Southern Avenue and Central Avenue in Phoenix. Our Operations Control> Center was immediately notified, light rail operations staff responded to> the scene, and Waymo was contacted. To minimize service impacts,> northbound and southbound trains exchanged passengers before reversing> direction to continue service. The incident caused no significant delays> and the scene was clear by 9:15 a.m.



Edited 1 time(s). Last edit at 01/09/26 09:02 by GenePoon.

I lived in the Phoenix area from 2017-2020. It seemed like there was a wrong-way human driver on the Interstate nearly every week. I believe they have set up sensors to detect the wayward humans and post warnings. Having recently been rear ended by a human driver (car totaled), I think I would prefer to trust Waymo.

Good job by the passenger who decided to save themself by getting out of the Waymo. Just because the car is dumb doesn't mean you have to be too.

Pretty scary, and funny too. What's next: Driverless trains that decide to go down the road, lol? Don't worry, technology and AI will have us all laying around soon, taking it easy while we're in the grave...

JG

GenePoon Wrote:
-------------------------------------------------------
> While incidents like this may draw skepticism
> about how safe these cars are, Maynard said he
> believes they’re likely safer than typical human
> drivers.

I don't fall for this zero sum argument. Why does it have to be human vs. machine? Is machine assistance of competent humans not allowed?

Waymo's problems here in the Bay Area in recent months: running into fences, and running over pets (smaller dogs).

I ignore the media hype, and overall, Waymo has made slow, but steady progress on their technology. Saying that is not sexy and controversial enough to satisfy the technology cult, skittish daytraders, or attention getters (readership on news and social media).

I personally don't "trust" self-driving taxis, but friends have ridden Waymos in SF recently after I showed them my videos of them running around. They are unhurt and still alive. :-)

Here's a little mind exercise. Would a robot riding Waymo escaped like the people in this story? It requires survival instinct, ability to improvise and anticipate ("imagine the future") to make that decision.

GenePoon Wrote:
-------------------------------------------------------
> While incidents like this may draw skepticism
> about how safe these cars are, Maynard said he
> believes they’re likely safer than typical human
> drivers. “These cars, in many circumstances,
> are safer than human drivers because they don’t
> have distractions, like a human driver does,”
> Maynard said. 

Which is a bunch of odorous excrement. It is not that human drivers are distracted, it is that human drivers make pathetically stupid choices to be distracted. 

Technology should be used to detect a human driver NOT paying exclusive attention to driving/operating a vehicle and bringing the vehicle to a stop if the driver is not doing what is required.

Quote from the article:

> > While incidents like this may draw skepticism
> > about how safe these cars are, Maynard said he
> > believes they’re likely safer than typical
> human
> > drivers. “These cars, in many circumstances,
> > are safer than human drivers because they
> don’t
> > have distractions, like a human driver does,”
> > Maynard said. 

Apparently, they DO have distractions, the railroad tracks in this case. Technology can greatly assist careful drivers (I love backup cameras), but I'd rather reserve judgement calls like railroad tracks to my own judgement. Your mileage may vary...

JG
 

This is really like pretty much any other new or evolving technology.  Because it is new and out of the ordinary it gets headline treatment.  A quick Google search tells me the Federal Transit Adiminstration has documented 2,316 collisions between vehicles and/or people and light rail transit vehicles between 2015 and 2023, averaging out to 289 a year.  Take the driverless component out of this incident and it might have been buried in the front section, if it even got reported on at all. 

Jeff Moore
Elko, NV     

JDLX Wrote:
-------------------------------------------------------
> This is really like pretty much any other new or
> evolving technology.  Because it is new and out
> of the ordinary it gets headline treatment.  A
> quick Google search tells me the Federal Transit
> Adiminstration has documented 2,316 collisions
> between vehicles and/or people and light rail
> transit vehicles between 2015 and 2023, averaging
> out to 289 a year.  Take the driverless component
> out of this incident and it might have been buried
> in the front section, if it even got reported on
> at all. 
>
> Jeff Moore
> Elko, NV     

Yep. Add Waymo or Tesla and bamm! Headline.

LongliveSP wrote:

"Technology should be used to detect a human driver NOT paying exclusive attention to driving/operating a vehicle and bringing the vehicle to a stop if the driver is not doing what is required."

Tesla does this now. Lots of complaints about driver monitoring in Self Driving mode. Some complaints are legit, but Tesla will work those out. Like seat to far back, some hats types and certain sunglasses. If a company's product did this full time, they'd sell just a handful and it'd be gone. Couldn't sell them. Only market being parents buying a car for their kids.

JDLX Wrote:
-------------------------------------------------------
> This is really like pretty much any other new or
> evolving technology.  Because it is new and out
> of the ordinary it gets headline treatment.   

Not quite, gotta add in the prospect for lots of deaths and destruction, and the public should be made aware. Already glaring errors and problems have manifested themselves, like in this case. Maybe they'll get worked out or maybe they won't, no one truly knows, although the companies and gullible public (including some TOers) tell us they automatically will. I reserve judgement...

JG

cchan006 Wrote:
-------------------------------------------------------
> Here's a little mind exercise. Would a robot riding Waymo escaped like the people in this story?
> It requires survival instinct, ability to improvise and anticipate ("imagine the future") to make that decision.

The car was never struck by a train. It was only human fear that led to the passengers fleeing. The robot rider would have been safe.

Not to say that I would not have bailed at the first opportunity. In hindsight it is easier to say "the riders would have been safe". The reaction of the train service helped keep that car safe.

For the robot to bail it would have needed to make a different decision than the vehicle as to the safety of the situation. The robot would have needed to recognize that it was in a vehicle on railroad tracks and that such a situation was a bad thing. From a practical standpoint I assume the robot would "trust" whatever vehicle it is in (the same way most riders trust public transportation) and only be concerned with getting in to the vehicle and exiting when informed that the vehicle had reached its destination.

Leaving a vehicle before the destination isn't "normal". The most I have done in real life is change cars on a train, although I could see changing trains if I felt the one I was on was unsafe (and service frequencies allowed me to catch the next train). I can imagine ending a taxi ride early if I felt unsafe - "just let me off here". But such a decision would be weighed against the hassle of getting another ride.

Yes, high level thought ... perhaps overthinking it. 

jgilmore Wrote:
-------------------------------------------------------
> Not quite, gotta add in the prospect for lots of
> deaths and destruction, and the public should be
> made aware.

Self-driving under a turning trailer? Running into firetrucks (always one color)? Or the city-wide power outage that disabled a bunch of robotaxis recently?

> Maybe they'll get worked out or maybe they won't,
> no one truly knows, although the companies and
> gullible public (including some TOers) tell us
> they automatically will. I reserve judgement...

I had the privilege of observing the evolution of Waymo vehicles the past ~10 years, watching strange evasive maneuvers, onboard technician interventions, intersection confusion and so forth. It doesn't quite jive with the financial media propaganda that have created a lot cult followers. But some things do get worked out.

As a research project, 10+ years for incremental improvements is OK by me. The rush to commercialization combined with Wall St. neurotic impatience (greed) is what's creating safety issues. Hint: don't tie your self-driving "advocacy" to your stock portfolio.

From my "tech experience," it's my opinion that some issues require very unlikely "breakthroughs" before they can be solved, which has NOT happened (i.e. "AI" is not the answer).
 

justalurker66 Wrote:
-------------------------------------------------------
> Yes, high level thought ... perhaps overthinking it. 

Appreciate your reply, contrasting human and robot reactions. The decision for the Waymo passenger to bail is interesting - I speculate that he thought the situation was spiraling out of control, so he bailed to mitigate risk.He didn't design the algorithm, nor knew the "flow chart" of robotaxi's decision process, so I can understand. A human Waymo technician likely would have acted differently.

We as humans have the luxury to overthink. However, we can often choose NOT to overreact. The safety-conscious decisions by the light rail "drivers" probably helped. Properly trained human beings.



Edited 1 time(s). Last edit at 01/10/26 18:00 by cchan006.

cchan006 Wrote:
-------------------------------------------------------
> Running into firetrucks (always one color)?

Firetrucks are always lime green now? Trucks are also available in teal green, yellow and red.

The biggest change Waymo needs is better customer service. Perhaps that has improved over the past couple of years, but it would show confidence in their own product if their rescue drivers would arrive in other Waymo vehicles (instead of self driven cars) or remote assistance solved nearly all problems remotely. The company still has more confidence in their system and AI than remote assistance providing a solution. (Remote assistance cannot drive the car, they can only make a suggestion to the onboard system. The onboard system remains in control of the vehicle and will not follow advice unless the AI of the onboard system agrees. Unless a rescue driver arrives and takes manual control all movement must be approved by the AI.)

In this case (Waymo on light rail tracks) serious reprogramming is needed to teach the Waymo cars NOT to go in lanes and roadways dedicated to light rail (the "I was following the GPS" excuse doesn't work for humans). It apparently stopped for the approaching train on the next track but was otherwise stuck in a situation that it should have been able to escape (or at least have been directed out of by the human overlords). Unfortunately when the company lets the AI overrule the human remote tech support the situation is more of "robot overlords" than "human overlords".

Perhaps hefty fines against Waymo and other automated driving system providers when something goes wrong would help? 

justalurker66 Wrote:
-------------------------------------------------------
> Firetrucks are always lime green now? Trucks are
> also available in teal green, yellow and red.

That wasn't Waymo that drove under a trailer or ran into firetrucks (some of those were fatal). I deliberately didn't mention which brand to avoid unnecessary TO rant fest, BUT it's worth clarifying to you. There's an age old optical sensor issue for recognizing certain colors - I've known about it for decades. Anyway, shortcut is to guess by software ("AI") but those damaged firetrucks are proof those earlier "fixes" didn't work.

The Arizona fatality (fatally running over a woman) was done by a Uber self-driving vehicle under testing, with a technician onboard. So far, despite issues, Waymos seem to have a decent safety record, except for the smaller dogs, one of which reportedly died.
 

cchan006 Wrote:
-------------------------------------------------------
> That wasn't Waymo that drove under a trailer or ran into firetrucks (some of those were fatal).

Your comment about fire trucks was not Waymo specific. My response about the various available colors is not Waymo specific.

The challenge of designing and operating AI controlled vehicles (whether it be Waymo or other developers) is an industry wide challenge. International standards have been set and Waymo refers to those standards when discussing the safety of their vehicles. It seems that every developer is using the same talking points.

In any case, all of the developers need to improve. Waymo is on the bleeding edge of driverless (no safety driver, no requirement for the occupant to be ready to resume control at any moment, fully automated driving). Other developers can blame the humans for using automation in situations where it should not be used or failing to monitor the automation and prevent the automation from causing an issue. Waymo seems to have taken the opposite approach relying on the AI more than human intervention.

I am glad that Waymo is not available in areas where I drive. I don't consider the system ready for use on public roads.