San Francisco bans delivery robots


San Francisco just passed legislation largely banning delivery robots on the city's sidewalks. The rule allows each company testing such robots to get a permit which would allow only a trivial number of robots, and limit them to industrial streets at low speeds for testing only.

I should note that I am an advisor to Starship Technologies, which is by a decent margin the leading company in this space. Starship has not conducted operations in San Francisco (other than a couple of manual-drive demos) and accepts the ordinance. I am not speaking on behalf of Starship in this article, however. Starship does have operations in Redwood City, south of San Francisco. Other companies, such as Marble, have been conducting operations in SF and are more affected by this ban. Because other cities are welcoming to the robots, it seems unlikely that any company not based in SF would conduct anything but limited early testing there.

This isn't the first time San Francisco has done a preemptive ban on a nascent technology. Famously, it also banned the Segway on sidewalks after most other cities voted to permit it. Nonetheless, it is at odds with SF's status as the "big city" part of Silicon Valley and the great beneficiary of its economic boom.

The reasoning behind the ban is quite flawed. Supervisor Yee, who is the primary author of the ban, expresses worry about safety and sidewalk congestion, stating that sidewalks should be reserved for people. The first is an error, the second has merit but not as much as would be suggested.

According to the DoT's household transportation survey, a whopping 45% of all trips are for shopping and errands, and 87% of trips are in cars. In other words, the use of delivery services has immense potential for reducing car miles, and with them, car accidents and injuries. Most crashes, injuries and fatalities are on city streets, not highways, in spite of the lower speeds. If a delivery robot takes a car (or delivery van) off the road, it's a win for safety. It's also a win for congestion, both from the real travel and from the "hunting for parking" which makes up a lot of urban shopping drives. Various studies calculate an average of about 30% of cars in shopping districts are hunting for parking.

Safety, congestion and energy

For rush delivery, delivery robots offer several huge wins. Compared to street vehicles, they use no road or parking capacity, and being small and light, they use far less energy per mile. All known delivery robots are electric, meaning emissions are also vastly less even for the same energy. They can thus do rush delivery at a much lower cost.

(I say rush delivery because a large delivery van full of non-rush packages can be efficient in terms of energy per package, though it still takes road capacity.)

Delivery robots are small and light and not very fast. The Starship robots are designed to be so light and slow that even if their safety systems were to fail and a robot were to hit a person, the chances of injury are very small. I've been hit by one deliberately and while you would not want it happening every day, it is no big deal. Some companies are making much heavier robots so this may not be true for all.

For cities, delivery robots also offer something desired by local retailers -- an opportunity to compete with the giants of online commerce like Amazon. Local stores, which keep goods close to customers, can offer faster service on many products than warehouse operations can.

With all this, it's unclear why a city would wish to ban a technology that improves safety, reduces congestion, lowers emissions and benefits local merchants. Of course, I have a bias, but the case doesn't even seem close.

Who uses the sidewalks?

The second factor, namely congestion on the sidewalks, and the idea of preserving sidewalks for people and walking, is more worthy of discussion. Delivery robots may not make as much sense on downtown sidewalks which are crowded enough to make this an issue. Most sidewalks, particularly ones in low to medium density residential and commercial areas, are highly underutilized. It is these areas that are the prime targets for the delivery robots. (Not the industrial areas which San Francisco's rule covers.) Delivery robots should be good citizens on the sidewalk, politely yielding to pedestrians and even ducking out of the way into available off-sidewalk space as needed. Robots are not in so much of a hurry that they have to hog the sidewalk, and well behaved ones should not.

The disabled

It should also be noted that delivery robots will be a boon to the disabled and elderly , who don't find it as easy to go out and do shopping trips. The ability to get most products right to your driveway in 30 minutes will allow people to continue to live in their homes when otherwise they might find themselves forced to move. Many of these people can't drive, and so they're the very people we want to preserve the sidewalks for, but they can make use of the sidewalk without traveling on it personally.

Ban before deployment?

Generally, cities and other governments should avoid regulating technologies before they are deployed. It is the job of the technologists to identify and solve the problems. Cities should not try to guess the problems in advance, and certainly should not regulate to stop things that have not actually ever happened. When companies show that they can't be trusted to act well, that is the time to regulate what they do. Nobody, neither city councils nor the roboticists developing these devices know the final form they will take, and nobody should cast any ideas in stone today. Once something is banned, it is very hard politically to "un-ban" it because nobody wants to be blamed if something they un-ban does cause a problem. San Francisco may switch from being a global technology leader to the backwater which is the last place to benefit from it.


I support the ban, actually. Sidewalks are for humans. Putting other vehicles on them will cause all kinds of issues (safety, congestion, etc). There may (may) be some benefits to robodivery vehicles but they will need their own infrastructure that is neither a bike lane nor a sidewalk. That’s not happening any time soon so maybe vacant suburban sidewalks are a better test zone to show impact. I have some concern though of everyone having 1 item delivered from every random store leading to more inefficiencies despite EV and other heralded car-reduction claims you cite. Would need more data to see how people really use it and impact it has, which I assume is basically what SF gov is saying- prove it.

Sidewalks are indeed intended for humans. But in many places the humans are barely using them. Most places. Why not make use of that to get cars off the road? The robots are only running to serve humans, and the real purpose of the sidewalks, and roads, is to serve the people. If a small robot cruises the sidewalk it's better than if the customer drove to the store to get the item, no?

If we get to the point that there are so many robots on the sidewalk that it interferes with people walking or jogging on them, I could see being against that The cities should reserve that power.

Thank you, Brad. I enjoyed your article and agree with it.

Add new comment