In Media Debut, Taco Truck Trips Up Autonomous Chevy Bolt

Chevy Bolt


Autonomous Chevrolet Bolt out for a spin (via Glenn L)

Though some automakers are working diligently on self-driving cars, this Chevy Bolt proves that it may still be awhile before a robot can fully comprehend its environment.

GM’s Cruise Automation initiated the first public rides in its autonomous Bolt this week. Reporters and non-employees were able to experience the self-driving vehicles first-hand during this official media debut in San Francisco. Previously, only Cruise employees were able to partake.

As the Bolt was making its grand arrival (which consisted of about two miles of slow travel through a thick crowd), it didn’t seem to go quite as GM had likely planned. Orange traffic cones were an issue for the Bolt, along with double-parked vehicles. However, the most interesting part of the story is the Bolt’s first encounter with a taco truck.

Chevy Bolt

Autonomous Chevy Bolt Testing In Arizona

These cars are tested rigorously to “see” and “understand” normal (and even abnormal) objects and situations on roads. We recently shared a video and story about Waymo’s self-driving Chrysler Pacifica Hybrid, which explained that the cars are tested in simulated and real environments with other vehicles, pedestrians, bicyclists, intersections, emergency vehicles, and even an occasional squirrel. However, perhaps not with taco trucks.

To put it into perspective, Reuters reported that during the 2.2-mile, ~15-minute drive, the self-driving Bolt happened upon 117 people, 129 cars, and 4 bicycles (this information was all recorded via the car’s sensors). According to Cruise CEO Kyle Vogt, the car didn’t exceed 20 mph and worked its way around:

” … urban traffic, a tram line, construction zones, pedestrians crossing streets and many double-parked vehicles.”

The self-driving Bolt is more cautious than most human drivers, as it slows almost completely for any of these obstacles. It waits and attempts to find a clear path. The cones and double-parked cars caused the Bolt to stutter a bit longer, but eventually, it made its way through. However, the autonomous Bolt stopped and “stared” at the taco truck (complete with pedestrians ordering lunch) for over a minute without proceeding. The human driver had to engage. Heck … this is much better than reporting that the Bolt slammed into the truck or the pedestrians.

Vogt said issues like this will improve over time. He believes that the winner in the self-driving car race will come down to which company can pull it off on a massive scale, rather than who’s first to market. GM President Dan Ammann added:

“Our mission is to bring this technology to commercial deployment at scale, with safety, as soon as we humanly can do that.”

Source: Yahoo News via Reuters

Categories: Chevrolet

Tags: , ,

Leave a Reply

32 Comments on "In Media Debut, Taco Truck Trips Up Autonomous Chevy Bolt"

newest oldest most voted

It’s clear what happened. Bolt took a toke before the drive, and got hungry when he (always a he) saw a taco truck. Pretty soon, we’ll be testing self driving cars for sobriety.

The car was just stopping for a byte.

That’s gold! Ahahaha


Or was the bolt just stopping for a bit?

The Bolt autonomous software likes soft tacos.
Obviously, this stop was on a “Taco Tuesday”, and all passengers agreed, “it’s definitely worth the wait!”


It would be interesting to know how much of an advantage Tesla has with its huge data collection efforts. If I understand correctly they can upload sensor data from the majority of the cars they have sold

No autopilot we’ve seen from Tesla is this advance. Cadillac’s SuperCruise is closer to Tesla’s current autopilot.

Of course the production version of autopilot enabled for customers to use today isn’t as advanced as an experimental test program.

The issue at hand is how much more data is Tesla successfully collecting running in non-autonomous mode that they will be able to leverage in the future for turning their future autonomous mode. From what I understand, this is the key to how they are developing their system, using hundreds of thousands of cars. Not small test fleets driving around generating much less data.

Tesla isn’t even in this race.

Tesla cant stop and go, and it will have the same issues with obstacles and pedestrian and moving white trucks

The average Subaru has a more advanced ‘autopilot’ than Tesla. Tesla will get there, but I wish people would stop claiming Tesla leads in this technology. It is interesting that the approach Tesla is taking will allow them to deploy their system to waiting hardware when the software is ready to go.

That assumes the sensors fitted into the cars are sufficient, which I’d wager isn’t going to be the case. They’ll probably be sufficient for highway driving, but not advanced enough to deal with the complexities of urban traffic.

With that being said, highway driving is of the greatest importance anyway, since it’s something that human drivers don’t particularly enjoy or handle well (distracted, fatigued, etc.).

How is this a flaw? I mean, who doesn’t like tacos!?!?! The car obviously assumed the passenger was going to want a taco…duh!

AI is farther along than I thought!


And one minute is the appropriate and polite length of time to wait and see if any of the passengers want tacos.

The car’s electronic brain was caught in an epiphany of wonder, awestruck at its first encounter with the elegant simplicity of a truck that converts into a taco stand. We humans have sadly become jaded, and ignore the magnificent splendor of food vendor trucks! 😉

I see a lot of Cruise Bolts here on the streets of San Francisco. I also see a lot of Tesla here as well. The Tesla’s are always being driven by their single passenger owners, they seem never to be driven in auto pilot mode. All the YouTube videos posted of Tesla’s in auto pilot mode are on freeways. So my hunch is the mass data Tesls has is of freeway driving not of driving in dense urban streetscapes. I suspect there is quite a difference between lane changing and off and on ramp freeway driving compared with turning from a two way The is ended straight ahead as it is becoming a one way street directly onto another one way street with dense cars and buses, bicycles, pedestrians, and robotic sidewalk devices all over the place that we have now in San Francisco.

Yes, Tesla does not enable autonomous driving in customer cars. Neither does any other car company, including GM.

Are you seriously expecting live customer cars to have the same features as experimental test cars like this one have?

I’m not sure the point of your comparison.

My point of comparison is, I do not see Tesla’s in data gathering, experimental mode driving San Francisco Streets. There are probably more X’s and S’s around me than Bolts. I do not hear of such tests in Palo Alto or around the Bay Atrea either. So is the massive Tesla data base just highway driving? How and when will city driving come to Tesla Auto Pilot?

When will they build an all electric, fully autonomous Taco truck? Not only self driving, but self cooking and dispensing.

Sir, you have higher aspirations than me. I only dreamt of a taco truck that doesn’t give people the runs.

Here’s the Reuters article:

A TechCrunch writer also took a ride. No human intervention, though the car did pause a while behind a cherry picker (tacos, cherries, …. hmmm):

Cruise pats themselves on the back for tackling a tougher environment than Waymo, but I’m not sure San Fran at 10-20 mph is truly harder than suburban streets at higher speeds. Waymo’s s/w is clearly more refined, and they’re launching much sooner. But GM is committed to mass production and might be able to move faster than Waymo once they get rolling (ha).

The Bolt is probably a better urban robotaxi platform than the Pacifica vans Waymo uses. It also gives GM a leg up with cities looking to set up no emissions zones.

The average EV owner wants a car that can do freeways autonomously. Nobody wants to do boring 2-6hr drives. It’s great to be able to get in my car, get on the freeway, have the car take over, I go to sleep, then get awaken at my destination. I can get off the freeway on my own, thank you.

Automated city driving benefits ride-sharing like Uber or Maven. I’d really, really hate driving behind a sub 20mph car that stops at every frigging little thing. My horn would have long overheated.

“He believes that the winner in the self-driving car race will come down to which company can pull it off on a massive scale, rather than who’s first to market”
So true, says the CEO whose specially-made $100K++ cars still look like a science project.

In the end, it’s a step in the right direction. Just think that some folks are too eager to pat themselves behind the back.

It was taco Tuesday. Totally understandable.

My name is jeff and i like trains. 21

It’s like when the audience crucified the judges for giving Olga Korbut less than stellar marks when she didn’t perform the routine no one else go do perfectly. Basically if you’re the only one who can do it you should get the most points.

As for questioning whether San Francisco at 20 MPH is more difficult than driving in Mountain View, that’s funny. Driving in San Francisco is like Pac Man. Mountain View isn’t comparable.

Fleet supposedly out by the end of 2019. Amazing how fast this is happening. The best aspect of all this is that GM claims that autonomous vehicles need to be electric vehicles. This means all manufacturers are going to have business reasons to develop EVs. Adoption won’t have to rest on government fiats.

Yep ahead of the competition by at least 2 years.

Actually this is pretty impressive and it would be fine for any normal vehicle. If the vehicle runs into trouble it just stops and beeps at you and you take over.

It isn’t fine for those who have some burning desire to no longer need a drivers license to use a car. Nor does it work for companies that want to do away with human taxi and truck drivers. But frankly, I’m not sure that’s necessarily bad for anyone besides Uber and Lift.

Perfect, maybe the Bolt can stop in front of the Taco Truck so I can call the Taco Truck for an order and then the Bolt can deliver the Taco back to my office when it is ready.

Stopping by Taco Truck is a MUST for me. Maybe GM will charge another $750 for that option.

The system needs to be overly cautious in this early stage of development. Any issue caused by the system no matter how small will be blown out of proportion by the media and that will hurt development and likely also cause overly aggressive regulations.

Stopped for a taco truck?

Maybe it needed gas?

“2.2-mile, ~15-minute drive”

That’s pretty slow. Could have biked it faster.