Autonomous Chevy Bolt Owned By Cruise Automation Hit By Motorcycle

Chevrolet Bolt


Chevy Bolt

Autonomous Chevy Bolt out driving

The trend continues … self-driving Chevy Bolt EVs keep getting hit by human-operated vehicles.

A few months back, we reported that these GM/Cruise Automation autonomous cars seem to be continual victims of human error.  According to a recent California DMV report (included below), it has happened again.

No matter how much self-driving car tech improves, there will always be misjudgments and mistakes by human drivers – often in anticipation of driving patterns of normal humans (over the rules of the road), which will cause accidents.

Having “robot” cars on the road along with people is a risky mix. Obviously, one of the number one reasons for the technology itself is that human error leads to an incredible amount of vehicle-accident-related injuries and fatalities. Until all (or at least most) cars have standard driver assistance systems, the problem will persist.

Chevy Bolt

Chevy Bolt EV autonomous test vehicles are assembled at General Motors Orion Assembly in Orion Township, Michigan. (Photo by Jeffrey Sauger for General Motors)

To be fair, we did also report a recent problem with the autonomous Bolt itself. However, it was minor and didn’t result in an accident. It turns out that certain “obstacles” are still not fully understood by the onboard artificial intelligence. Basically, the Bolt was taking it very slow due to a bunch of parked vehicles and hesitated too long when it happened upon a taco truck. Eventually, the human driver had to intervene to move the car along.

We’d much rather have our self-driving car stop and wait in confusion than plow through a crowd of hungry pedestrians. Even if they weren’t hungry, it would not be a positive situation for the eventual deployment of the systems.

Fast forward to last week in San Francisco, and we learn that an autonomous Chevy Bolt owned by Cruise Automation collided with a motorcycle. Again, it was not the fault of the self-driving car, and fortunately, the motorcyclist walked away with only a sore shoulder. Cruise explained:

“… a motorcycle that had just lane-split between two vehicles in the center and right lanes moved into the center lane. [The motorcycle] glanced the side of the Cruise AV, wobbled, and fell over.

We test our self-driving cars in challenging and unpredictable environments precisely because, by doing so, we will get better, safer AV technology on the roads sooner. In this case, the motorcyclist merged into our lane before it was safe to do so.”

The Bolt was only moving at 12 mph and the motorcycle was traveling at 17 mph. Cruise CEO Kyle Vogt shared that San Francisco is the perfect location for testing these vehicles since difficult situations occur more often than most other locations. He wrote:

“Our vehicles encounter challenging (and often absurd) situations up to 46 times more often than other places self-driving cars are tested.”

From September through November, Cruise reported 14 incidents to the California DMV. Click here to see the recent incident report as filed with the California DMV.

Source: Ars Technica

Categories: Chevrolet, Crashed EVs

Tags: ,

Leave a Reply

48 Comments on "Autonomous Chevy Bolt Owned By Cruise Automation Hit By Motorcycle"

newest oldest most voted

This should make it through mainstream media as “another autonomous car crashes”.

It was reported pretty widely two weeks ago. Every article I saw said “motorcycle hits GM autonomous car” or similar.

The motorcyclist was ticketed, but the articles make it sound like the Cruise AV was just minding it’s own business. It actually caused confusion by starting to change lanes then moving back into its original lane unexpectedly.

AVs can contribute to wrecks by doing things that are legal but different than what a normal human would do. Waymo spends a ton of time getting their cars to act more “human” to better blend in with traffic. They got rear-ended a lot in early days because they’d do stuff like start to make a right turn then stop suddenly out of an abundance of caution because they detected a distant car.

There’s not enough detail in this report to know, but it says the Cruise AV aborted the lane change when a minivan in the new lane they slowed down. I’d bet money this was an “abundance of caution” move. The motorcyclist, seeing the Cruise AV changing into a lane which had ample space, never imagined it would suddenly reverse course. He aimed for the newly-vacant space only to find it unexpectedly become un-vacant.

This is probably the best post I’ve seen to respond to “another autonomous vehicle crash.” Autonomous vehicles do not drive like humans and therefore they do cause accidents. Thank you for writing it.

I wonder if the autonomous vehicle program has a lane-splitting algorithm. My guess is that there is a whole lot of code missing. How about “snow and road salt so bad that the lane markers are invisible?” or “handling for black-ice conditions?”


Well written!

Accidents are caused by idiotic drivers (be it humans or AI). Drive smoothly and others around you can anticipate and react accordingly. I ride a motorcycle and that Bolt doing that would have gotten the finger from me.

Even slow drivers (with abundance of caution) can cause accidents. By going slower than the speed of traffic, they cause others to weave around them, maneuvers that increase the risk of accidents.

I’m in SF every now and then and frequently see these idiotic AV’s impeding traffic.

Quote from article:
“We’d much rather have our self-driving car stop and wait in confusion than plow through a crowd of hungry pedestrians”
Please stop stopping every 5 seconds. It is terribly annoying and dangerous. If you’re stopped when you don’t have to, people behind you will try to get around and you’re blocking their visibility (in addition to their progress).

I disagree with the front car causing the rear car to rear end it just by stopping or slowing. Law says rear car is always at fault. People think they can take right turns at full speed, ignoring peds, bicyclists, etc. An ounce of caution makes plenty of sense.

Brian said:

“Law says rear car is always at fault.”

“If the law supposes that… the law is a ass — a idiot.” — Charles Dickens, Oliver Twist

If a driver, either human or robot, suddenly performs an unexpected maneuver in dense traffic, then he may be the cause of the accident, even if he is hit by a following vehicle… regardless of what the law says.

I understand why the law is written the way it is; it helps settle disputes, and it’s right more often than not. But to claim it’s one person’s fault just because “the law says so” is a rather silly argument. A vehicle approaching another vehicle from behind may indeed always be “at fault” according to legal liability, but that’s rather far from being equal to actually being the cause of an accident, in the logical or scientific sense of cause-and-effect.

Computer programmers who write code for controlling autonomous vehicles must follow the dictates of logic and the physics of moving objects dictate. The “rules of the road” are not necessarily what is written into the legal code!

No, pushmi. The onus is always on the following driver to leave the correct gap to the vehicle in front, and stop in time. Just because a vehicle in front stops unexpectedly when you’re anticipating they’ll carry on, it’s STILL your responsibility, as the following vehicle, to give them space and not accelerate unless it’s safe to do so.

If you seriously think the law protects idiots, then I dare you to slam on your brakes in traffic. Then if you get rear ended, go blame that driver.

You’ll be lucky to walk out of that alive.


Hey, that’s a fantastic write-up of the problem! Doggy, may I have your permission to copy your post to start a discussion on the InsideEVs Forum site?

* * * * *

Vexar said:

This is probably the best post I’ve seen to respond to “another autonomous vehicle crash.”

It is certainly the most informative that I’ve seen on the subject, and also very well written! Thanks to Doggydogworld, I learned something today. 🙂

So you’re suggesting that the driver of the motorcycle was monitoring the lane in the other side of the Bolt but not monitoring the Bolt itself closely enough? You infer that this overly cautious Bolt suddenly and erratically veered back into it he lane it hadn’t vacated. That’s a dubious inference.

A car aborting a lane change IS a common traffic event. There is nothing in the report that indicates the Bolt behaved unlike a good driver.

He was on a Honda S90 which is much more responsive and maneuverable than a car. In this case the motorcycle was traveling faster and struck the Bolt and was cited for his failure.

What is “glanced the side?” Was the biker glancing (looking) at the car or did he make contact?

Glancing blow, side swipe. basically scraped the side of the Bolt wobbled and fell over.

Read the accident report PDF. Still not clear what “glanced” mean.

It seems Bolt started the lane change, but then went back into its own lane, perhaps suddenly after sensing space is too narrow due to car in front on next lane slowing down. If human driver, he’d most likely slow down rather than go back to his own lane, thus causing everyone to slow down. I think the motorcyclist probably anticipated such human behavior.

Yes, technically Bolt wasn’t at fault, but it was not following typical (aka, poor) human driving behavior.

It wasn’t a direct hit, it scraped the side, hence Glace. It’s a fighting term.

I don’t remember seeing fighting terminology in my drivers training manual.

“Read the accident report PDF. Still not clear what ‘glanced’ mean.”

It must mean sideswiped; i.e. “glanced off”. It can’t possibly mean merely that the cycle rider put eye-tracks on the Bolt; that wouldn’t be relevant to an accident report.

There is an issue with a driver’s (or AI) interaction with motorcycles. For example, in HOV lanes in California, Motorcycles EXPECT cars to move over and make room for them as they split traffic. It is not written in any laws that they have the right of way and in fact they are supposed to yield since they are approaching from the rear, but it’s just something the motorcyclists insist upon.

There are many other examples that come into play, but the point is that since motorcycles don’t follow the letter of the law when it comes to rules nor are they consistent in every instance, it makes it difficult program for all variables.

I can tell you don’t ride. Motorcyclists with more than a month of riding do not insist on cars doing anything special when splitting lanes. If anything, cars making sudden wide berth on our approach make us nervous as do some a-holes purposely blocking us.

As for letter of the law, lane splitting is legal in CA as is riding in carpool lane. What law are they not following?

I didn’t say that there is anything illegal about splitting lanes, but I do notice daily that there are motorcyclists that give the finger to people who stay centered in their lane (equal distance on the left and right).

In other words, if they don’t squeeze to the left side of the lane, the motorcyclists will at the least, gesture to move over and at worst pull in front of those cars and stop.

“21750. The driver of a vehicle overtaking another vehicle or a
bicycle proceeding in the same direction shall pass to the left at a
safe distance without interfering with the safe operation of the
overtaken vehicle or bicycle, subject to the limitations and
exceptions hereinafter stated.”

You wrote “since motorcycles don’t follow the letter of the law”. Not following the letter of the law is illegal.

Yes not following the letter of the law is illegal.

Then what law are motorcyclists not following?

I’ve seen how many Americans lane split. It’s far more reckless and crazy than the filtering technique we’re actively taught in Europe. Besides that, over here most drivers do move over to give you room, so it doesn’t make you nervous. It makes you grateful.

Motorcyclists, especially young ones, are a large source of donor organs. I’ve driven with two ER physicians who refered to them that way. I am horrified by what I see them do on the highway, especially the lane splitting.

They are legal and will certainly stay that way but they will remain dangerous to themselves and others.

The crash in the article was not due to lane splitting. If the motorcycle was on center and thought the car ahead was going to change lane like a human driver, it would’ve resulted in exactly the same sort of crash.

Lane splitting is in some ways safer than staying in center. Distracted drivers often rear-end motorcycles if staying in lane in slow traffic.

By far the biggest hazard to motorcyclists are cars turning left in front of them, accounting for over 40% of all motorcycle crashes involving cars. If the ER doc actually knew the data, he would’ve said “organ donor generator” to every car turning left, because cars cause vast majority of motorcycle crashes.

But of course, like most of us, he’s completely ignorant of facts outside of his specialty, just like Nobel Prize in Chemistry winner Linus Pauling prescribing vitamin C for colds.

I’ve had cars pull out right in front of me. Luckily my safety responses kicked in subconsciously, before I had time to think.

I usually just pretend I’m invisible when I ride, and expect the worse.

The exact details are kind of irrelevant. This is precisely why I’d never ride a motorcycle. It hardly matters much who’s at fault if you’re dead.

A four wheel vehicle with air bags and a steel safety cage is going to get you through worse accidents than a tiny two wheeled vehicle with absolutely nothing.

Not going outside of your house is whole lot safer than driving. By your logic, you wouldn’t dare drive your car to a park or to see a movie or for any kind of recreational activity. Heck, you should just stay home and order everything delivered to you and work from home instead of stepping out the door.

At some point, you choose to live life, not imprisoned by it.

“This is precisely why I’d never ride a motorcycle.”

Good. A man’s got to know his limitations.

Your choice to not ride a bike, but you’re missing out on a whole world of fun. If you think cars give you a sense of freedom, a bike takes that feeling to a whole new level.

I admit I’ve only been riding motorcycles for thirty years, but I don’t “expect” cars to do anything except the unexpected.

Lane splitting (or filtering) is illegal here on the East Coast, but even if it wasn’t, I’d rather not do it because cars driven by humans are unpredictable. But that doesn’t mean that other riders don’t do it, you should see the Belt Parkway, it’s practically a way of life for them.

Now, weaving on the other hand…

In the UK your driving lessons include filtering; also on your license test if you don’t filter when it’s safe and legal to do so, they might even talk to you about it.

Here’s how this most likely play out, a slew of law firms will reach out to the motorcyclists, law firm will reach out to Cruise/GM demanding a settlement or will sue publically…Let’s see how this plays outs…

“…San Francisco is the perfect location for testing these vehicles since difficult situations occur more often than most other locations”

I that a politically correct way of saying that SF is full of a$$hat drivers?

It must be hard for the Autonomous Car programmers to balance “follow the law” and “drive like people expect”. The former is safer from a legal standpoint, the latter safer for the general public.

Autonomous driving or not, lane splitting should be banned. It sounds good in theory: If car traffic is stopped or almost stopped, then allow motorcycles to slowly and safely filter between them. In practice, it’s turned into motorcycles inventing dedicated, high speed motorcycle lanes. If car traffic is going 35, then they’ll race between cars at 70, flying into road rage if a car moves so much as one millimeter from the exact center of its lane. It’s ridiculous. Car drivers cannot be expected to constantly drive as if their twelve foot wide lane is eight feet wide.

Banning is a knee-jerk reaction. Filtering has been legal where I live for decades and 99% of bikers do it very responsibly. Why? Because when we take lessons for our license most instructors will teach us how to filter safely. We don’t attempt it when traffic is already flowing nicely, and if we do attempt it we make sure 1. we’re visible and 2. we’re not going more than 5-8 mph faster than the traffic we’re passing.

besides which, if you ban it people will do it anyway. bikes sitting in traffic is stupid when it just adds to congestion for EVERYONE.

San Francisco Streets are a challenging place to drive not so much because of the drivers but because of:
.intense traffic congestion, in some areas lasting many hours a day
.a street pattern that has is not a simple Cartesian grid but has heavily trafficked major diagonals
.a tweaked traffic flow pattern in which apparent two way streets change to one way streets and then back to two way along the length of a street.
.extensive use of Uber and Lyft cars that stop anywhere on a street including in the middle of a street to pick up and drop off passengers. (This is the home of Uber by the way.)
.major street and building construction going on all the time.

So for instance most Market Street intersections have 3 streets (6 direction) coming together. At certain points along Market streetStreet, all non commercial vehicles are turned off allowing only buses and trucks and cabs to continue. The AI to learn this seems much more complicated than that for freeway driving. The signage and barricades that signal these traffic flows go well beyond a light w red yellow and green.

Q- “Where am I?”
A- “You’re in a Johnny Cab.”
Q- “How did I get here?.”
A- “The door opened, you got in.”

I don’t Recall this… Totally.

If this keeps happening a lot more frequently than is the case for other self-driving car fleets, which seems to be the case, then I’m guessing that the cars in question are stopping more abruptly than human drivers expect. Perhaps a lot more abruptly.

If that’s the case, then why the heck haven’t Cruise Automation’s engineers fixed the problem?

Color me unimpressed.

That’s a lot of presumption. What’s missing in articles like this are comparable numbers. Like what is the accident rate based on How many miles per incident driven for a typical human operated cars (how many small fender benders go unreported because no real damage, and drivers don’t want to bother with insurance companies). Vs the same statistics for the autonomous vehicles?

It would indeed be good to see a comparison of the miles traveled by the various test fleets, along with the number of accidents they are involved in.

But I’m rather skeptical of the repeated assertions by Cruise Automation that San Francisco is somehow a significantly more difficult and dangerous place to drive than other cities. SF isn’t the only city that has large sections of streets not laid out strictly in a grid, nor the only city to have streets which switch from one-way to two-way traffic, nor the only city with narrow streets, nor the only one with high density traffic.

Perhaps SF is worse than most other places, but if so, then I would imagine it’s merely a difference in degree, and not a fundamentally different kind of city, as as Cruise Automation seems to be claiming.

Let’s look at the insurance rates for SF as compared to other cities; that would be the telling data, as the actuarial tables insurance companies use for insurance rates are based on actual accident statistics. According to the article linked below, San Francisco isn’t even in the top 30 most expensive cities for car insurance!

Nope, Cruise Automation: I’m not buying it.

How fast is the Bolt’s AI processor? Can it process LIDAR data, make a decision, drive smooth and safe at the same time?

Apparently it can.

It’s not all about the processor speed. For example, LiDAR gives a simple and direct way to measure distances to obstacles and to “paint” the shapes and outlines of the environment around the car, in real time. That requires much, much less processor power and number-crunching than does Tesla’s approach of relying mainly on camera images and complex, relatively slow (in processor speed) optical image recognition. LiDAR is a lot more reliable at spotting objects and obstacles, too.

As has been said many times by many observers: When building up a real-time virtual reality image of the environment, active scanning with LiDAR is far superior to passive scanning using camera images. LiDAR reflections also give a much, much sharper resolution; radar images are surprisingly fuzzy by comparison.

Poor reporting. No stats provided on how often human drivers hit other human drivers be how often AVs are hit by human drivers. Is it only news because an AV was involved?