Cruisin' for a bruisin': Deconstructing the first 'Level 4/5' AV-motorcycle accident

By now most HAVstory visitors already know that, about two months ago, a GM Cruise autonomous vehicle had a minor accident with a motorcycle on the streets of San Francisco. Although I’ve written elsewhere about other cars hitting motorcycles while in semi-autonomous mode (e.g. Tesla Autopilot) as far as I know, this recent Cruise incident is the first reported collision between a fully autonomous auto and a motorcycle on public roads.

 Last December, a Chevy Bolt test mule, like one of these, was involved in what I believe was the first collision between a fully autonomous vehicle and a motorcycle on public roads.

Last December, a Chevy Bolt test mule, like one of these, was involved in what I believe was the first collision between a fully autonomous vehicle and a motorcycle on public roads.

So, what happened?

The motorcycle rider, a Swedish photographer named Oscar Willhelm Nilsson, has initiated a lawsuit against General Motors. In the unlikely event this matter goes to trial, I suppose we’ll learn exactly what happened, but in the interim, based on the disengagement report GM filed with the State of California, the Allegations of Fact in Nilsson’s lawsuit, and email exchanges with both Mr. Nilson and his lawyer, here’s what HAVstory currently knows.

 This is a Google street view image of the block where the collision occurred. (No, I did not park that rare Honda Hawk GT in the frame just to make motorcyclists drool.)

This is a Google street view image of the block where the collision occurred. (No, I did not park that rare Honda Hawk GT in the frame just to make motorcyclists drool.)

The accident occurred on Oak Street, in San Francisco, just past the intersection with Fillmore, in the Lower Haight neighborhood. Oak is a one-way street with three travel lanes; it’s lined with parallel parking on both sides. It occurred on a Thursday morning, just after 9:30, at a time when all three travel lanes were occupied.

 Oscar Nilsson was riding this sweet little 1969 Honda S90. (It was misidentified as a 1996 motorcycle in GM's report.)

Oscar Nilsson was riding this sweet little 1969 Honda S90. (It was misidentified as a 1996 motorcycle in GM's report.)

The Cruise test mule was a 2016 Chevy Bolt EV, operating in fully autonomous mode with a test driver, Manuel DeJesus Salazar in the driver's seat . According to GM’s disengagement report, it was traveling in the center lane at 12 miles per hour, when…

“Identifying a space between two vehicles (a minivan in front and a sedan behind) in the left lane, the Cruise AV began to merge into that lane. At the same time, the minivan decelerated. Sensing that its gap was closing, the Cruise AV stopped maldng [sic] its lane change and returned fully to the center lane.”

While that was happening, Oscar Nilsson was filtering up through the Oak Street traffic at about 17 miles per hour. Oscar was riding a stylish 1969 Honda S90. He was ‘lane-splitting’ or filtering between lanes 2 & 3, i.e. approaching the Bolt from the right rear.

When the Chevy began moving to the left, Oscar also tracked forward and left – as you would when filtering to optimize the distance between vehicles. He moved into the space vacated by the Bolt.

It would be reasonable–and indeed a motorcycle safety ‘best practice’–to accelerate into that vacated space. For safety reasons a filtering motorcyclist always needs to be aware that any new gap can be seen as an opportunity for surrounding car drivers. It’s safer to be in that gap and ideally abreast of adjacent autos, rather than in the traffic seam, between that gap and a potentially inattentive driver.

But in this instance, when the CAV ‘changed its mind’ and “re-centered” in the middle lane, it sideswiped Nilsson.

This is where things get interesting.

Apparently, as reported in Traffic Collision Report #l70989746, the motorcyclist was determined to be at fault for attempting to overtake and pass another vehicle on the right under conditions that did not permit that movement in safety in violation of CVC 21755(a). (“The driver of a vehicle may overtake and pass another vehicle upon the right only under conditions permitting that movement in safety. In no event shall that movement be made by driving off the paved or main-traveled portion of the roadway.”)

Most California motorcyclists would not consider Mr. Nilsson’s maneuver to be an overtake. Although lane-splitting was legally defined in California in 2016, the practice continues to exist in a gray area. According to the California DMV…

California law does not allow or prohibit motorcycles from passing other vehicles proceeding in the same direction within the same lane, a practice often called "lane splitting," "lane sharing" or "filtering."

I traded emails with Sergei Lemberg, Mr. Nilsson’s lawyer, who wrote...

“Absolutely nothing prohibits safe and reasonable lane splitting at present. We believe the GM car was at fault, veering into and hitting Mr. Nilsson. The operator of the GM vehicle states in the police report that he saw Mr. Nilsson before the collision, but didn't have enough time to grab the wheel. The maneuver by the autonomous car was unpredictable and dangerous.”

It seems to me that the lawsuit, if it comes to trial, will hinge on a question of fact: Was Nilsson making an unsafe pass, or was he lane-splitting in a manner long accepted in California–which might mean the Cruise AV was at fault when it returned to its lane?

An alert human driver in that situation should have realized that when he started to vacate his lane position, another vehicle might move into it.

Considering Mr. Nilsson’s slow closing speed, a skilled human driver checking rear view mirrors every few seconds would likely have been aware of Mr. Nilsson’s approach in the 2-3 seam–but even if the driver was not aware of Mr. Nilsson, any vehicle in Lane 3 might have decided to move into Lane 2. If I got part way through a lane change to the left and suddenly changed my mind and decided to abort that maneuver, I would definitely throw a right-side shoulder check before moving back.

At the very least, the fact that the Cruise AV moved back into a space occupied by a full-sized person and a Honda S90 suggests an inadequacy in the Bolt’s sensor package. And Cruise got a valuable data point on the behavior of filtering motorcycles.

It’s interesting to speculate on whether, if this lawsuit goes to trial, GM will attempt to present video evidence from the Cruise AV. I imagine such video exists (and I would love to hear from anyone who can confirm my guess that it does.)

I ride motorcycles in California quite often. Frankly, I’m a little worried about a well-funded and motivated defendant (GM) establishing a precedent that would more strictly define lane-splitting and the conditions if any when it should be allowed. In fact, I’m a little worried that AV companies will put pressure on California to come into line with other states and ban the practice. The reason GM shouldn’t do that is simple: lane-splitting is an accepted and fully legal practice in the rest of the world. We motorcyclists enjoy the right to filter and AVs need to be able to safely deal with us.