In a short period we saw 3 at-fault accidents involving robocars (with one being purely the fault of the safety driver) and we're going to see more. We're going to have to learn how to deal with them, to tell the difference between serious error that says a team has deployed too early, and the accidents that will happen with miles because perfection is not a possible goal.
The future of computer-driven cars and deliverbots
I have often written about the debate between the robotaxi vision of self-driving and the private car sales vision. That debate got writ large last week with the firing of Cruise CEO Dan Ammann over his desire to push the robotaxi vision (and some other differences of view.)
I write about it on Forbes.com in GM CEO Mary Barra fires Cruise CEO over robotaxi/car sales battle
Mercedez-Benz has announced approval of their “Drive Pilot” system, in Germany. Tesla, on the other hand, doesn't do this because of their focus on the far-off goal of a "full" self-driving product. What does a traffic jam pilot really mean, and what could Tesla be doing if they weren't putting so much focus on the still far-off FSD?
I consider this my my new Forbes site article Mercedes Gets Approval For Traffic Jam Pilot, Where Is Tesla?
How can we tell how far along a robotaxi project is? They don't let us look under the hood, so we have to observe their real bets and milestones.
I've made a list of a rough order for the milestones. Most teams have far to go.
See Milestones of a robotaxi business at Forbes.com
In an earlier article, I noted that Cruise, in demonstrating their first robotaxi rides with no safety driver, did all the pick-up an drop-off by just stopping in the lane (late at night.) This is something many Uber drivers do as well, but it's not technically legal. Cruise is doing things one step at a time, but the SF MTA doesn't like that and filed an opposition to them getting a permit to operate the service with the public (currently they just do employees.)
Here's a Forbes.com article on the issues with doing pick-up and drop-off.
Two big milestones for Cruise this week, with two stories:
First, they started unmanned operations at night in San Francisco, and give their first taxi ride with no safety driver to founder Kyle Vogt. GM employees are now using Cruise vehicles as taxis.
Most of the major players want to run a robotaxi business -- Uber style ride service using robocars. Yet some have started to wonder if this is the best business model, or if it's even a good one, while companies invest billions in it.
In this new article on Forbes.com I investigate some of these questions and why the players are investing these sum, and what sort of profits they might make.
A video shows a WeRide safety driver apparently sleeping on the job on Highway 85 in San Jose. After Uber's fatality 3 years ago, are some operators still not monitoring driver attention?
I asked WeRide and learned only part of the answer at WeRide safety driver caught napping -- why is this still happening?
You have probably seen many demonstration videos of self-driving cars navigating the roads with aplomb. They show us a little about what the system can do but as long as they are cherry picked, they don't really tell us how the team is doing.
They could do better if they drive a random road at a pre-announced random time and stream it live, so there can be no cherry picking. Time to start.
NHTSA is investigating 12 crashes by Teslas on Autopilot into emergency vehicles on the side of the road. It's also asking the other companies who make products like Autopilot for their statistics. What can be done to prevent these crashes, and are any number of them acceptable? Is Tesla doing things wrong or doing it better than anybody else? We may learn that and the issues are complex.
I discuss them in this new Forbes.com article:
Companies are at the stage of announcing real pilot projects for robotaxi service. Now MobilEye announces they will start a robotaxi service in Munich and Tel Aviv by 2022. What are the new metrics of success for a team?
See more at MobilEye announces plans for robotaxis in Munich
The instinct of many transportation planners is to make "smart infrastructure," and to try to make plans for it going out 30 years. That's impossible, nobody knows what smart will mean in 5 years. The internet solve this problem, and grew by making the infrastructure as stupid as possible, and it revolutionized the world. The internet teaches lessons for how all infrastructure planning must go in the future -- keep the physical as simple as possible, do everything in the virtual, software layer.
In the robocar world, everybody is safety-obsessed. But what if what's holding things up isn't that, but the fact that focus on safety had delayed the good road citizenship needed to operate a real service. Is good road citizenship even harder than safety? What ways might we measure it and get the trade-off right. I discuss this in a new Forbes site article seen in:
Updating a few stories reported before (including Monday) I note that Scale.AI has launched their mapping service, which helps people tag and label maps more efficiently. Pony.AI shows off they are doing taxi tests with no safety driver in China, where roads are more complex than in Fremont, California. And Waymo shows off its Pride.
Read about these in this Forbes.com article Scale Mapping; Pony.AI driving in China; Waymo Pride
Last week saw a flurry of robocar news. The most significant was the deployment of no-safety-driver testing by Pony.AI in Fremont, CA, but there's also big funding news for Waymo, Cruise and Kodiak, Deepmap is sold to Nvidia, new taxis for Baidu, deals for Nuro and a collision between a Waymo and a scooter in SF -- in manual mode. Read all about it in my round up at:
Two major virtual self-driving conferences scheduled themselves for June 9. So I went to both, of course. Interesting news tidbids came from Argo, Starship, Scale, Chinese robotaxi makers, Zoox and many others. I summarize it here:
Cruise has been granted a permit to begin unmanned robotaxi tests in California. But they aren't allowed to charge, which forbids a lot of useful research, and explains why Waymo hasn't bothered with that permit yet. I discuss why it's so useful to experiment with payment in this Forbes site article at:
For a long time, Tesla refused to implement driver gaze monitoring with Autopilot. Now, quietly, the new release does it, though with no other changes as yet.
But it's quite a shift in their actions.
See the details at this Forbes.com story at