Two recent proposed regulations for robotaxis have come forward in DC and New Jersey, each supported by a different robotaxi player. Both laws, unfortunately, work against the public interest and have the appearance of attempts at regulatory capture.
New Jersey’s Bill S1677 doesn’t yet contain the language, but according to Wired, an Uber lobbyist proposed language that in order to offer robotaxi rides, a service would need to have human drivers provide 85% of its rides during the 3-year trial period. That essentially says that only Uber or Lyft could enter the game, which would be pretty direct regulatory capture. While it can be couched as job projection, it’s a bit like a law demanding that Ford motor, in order to sell cars, must also sell horses and operate livery stables.
Reports also stated a version of the law (not the current one) required that the cars have 3 different sensing modalities, ie. forbidding the Tesla approach of using just cameras. While I am a big booster of LIDAR and frequent critic of Tesla’s camera-only approach, mandating this view into law is foolish. The law should never state how to be safe, it should only require that the vehicles meet suitable safety standards. It’s up to the engineers to figure out how to do that.
Fortunately both clauses are not present in the current draft. Maybe the message is getting through.
DC Law
A law proposed in DC includes a very expensive permitting process (A $1M application fee and a $5M permit fee) as well as $5M of insurance and a $0.15/mile operating tax. Waymo is reported to be supporting this law.
These high permit fees pretty much exclude any small player from the game, particularly if you imagine needing such fees in every city or state. To be fair, today it’s not really possible for small players to make robotaxis, but that won’t always be true. Tesla has often stated plans to sell its “CyberCab” vehicles to individuals, though at present, if they wanted to then operate them as robotaxis they would need to do it through Tesla’s network, not on their own. That could change in the future, but not under a law like this. (Tesla’s CyberCabs do not currently work as safe robotaxis.)
I have written before about how Tesla’s private sale of CyberCabs is unlikely to work for technological and business reasons. But even if that’s true, it should be the business reasons that stop it, not because it’s illegal due to high red tape costs. My prediction could be wrong.
$0.15/Mile
The per-mile tax is wrong at many levels. Once you put in a tax like this, it’s very hard to get rid of it. As planned, this tax would not apply to other vehicles, so it ends up punishing the robotaxis. Robotaxis can be some of the best vehicles on our roads, solving many of the problems. The last thing you want to do is punish the good actors for being good.
Today, operating a robotaxi is expensive. Companies sell service for $2-$3/mile and lose money. That won’t always be true, and in fact the cost should some day get a lot lower. A low-cost single-person electric robotaxi, built for $10,000 at scale could have operating costs as low as 15-20 cents/mile. This tax could double that, which would be quite extreme. If there is to be such a tax, it should definitely have a sunset, or it should be capped at some fraction of the total sale price of the ride. (Except in future, robotaxi rides will be sold for flat fees rather than per-mile rates, making a per-mile tax a major impediment.)
There are valid argument that there should be a cost for all vehicles to use the roads, not just robotaxi. Today, that’s done with the gasoline tax, and already some complain that electric cars (which all robotaxis are) avoid this fee, and that’s unfair. It is, though today there is a strong public interest in promoting EVs, and the gasoline cars avoid paying for the huge externalities of pollution, soot and greenhouse gasses that come from burning their fuel.
Nonetheless, we should want all vehicles (even transit vehicles) to pay for the costs of the fuel they burn, for the amount of road space they take up, and other externalities like noise, congestion and crashes. They do pay for crashes, and some parking, though parking is often given free. Pick-up/Drop-off space is also given free, though it is used extremely briefly compared to parking space. In an ideal regime, the pricing would end up fairly complex. Road use would be charged for every square-foot-second used, at a rate which varies based on levels of congestion on the roads. Parking should be cheap, but never free, also based on the square-foot-second used and value of the location.
That, and other factors would be easy for robots to handle and calculate, but much too complex for human drivers. They prefer simple systems, like a flat congestion fee to drive into downtown, or a fixed fee for a day of parking. It’s possible for both to coexist, and for a road user to pick the system they prefer.
With the right rules, robotaxis will be among the best actors. They will carry passengers 50% of the time, while private cars are parked 95% of the day. Robotaxis never really “park,” they are actually “standing” and able to move as needed at any time. This will make them good actors, whom we want to punish the least with fees. On the other hand, all taxis must travel “deadhead” miles on their way to pick up riders, or driving back to the popular areas so as to be closer when they get a ride request. This creates extra burden which they should pay for. Because robotaxis don’t mind waiting, they will be better at avoiding deadheading in congested areas if that comes with a cost, as long as using uncongested areas has much less cost. We don’t want a simple flat fee. The robots can handle any amount of complexity in the pricing and will optimize their travel to reduce cost and thus meet the public interest as encoded in the fees. (Even human drivers, using tools like Waze, can be made to drive routes based on complex calculations, though humans care much more about the value of their time than robots do.)
Parking for robotaxis should become very cheap. I’ve predicted what I call the “spot market” in parking, a dynamic real-time bidding marketplace where both people and robotaxis can quickly find, reserve and pay for just what they need. Since robocars can park valet-dense and leave when asked, they will pay very low prices. Since they mostly want to be at work driving people, they also park a lot less, which means we need a lot less parking, and we can stop filling our streets with parked cars too.
