Using electric school buses to power the grid / Remote driving and Starlink


Vehicle to Grid (v2g) to provide power from car batteries is tough. A new venture wants to do it with electric school buses, which follow a fixed schedule and have big batteries. I examine how that would work at:

Electric Schoolbuses and V2G

Previously, I reported on Halo, a new service providing a car delivered to you via remote driving. You drive it yourself, a form of what I called a whistlecar many years ago. I discuss the advantages of that and debate if a network like Starlink could enable remote driving.

Halo Uses Remote Operations To Offer Summoned Carshare; Might Tesla Do It With Starlink?


It seems like school buses suffer from 2 obvious drawbacks for this purpose.
1. There are just not that many of them.
2. As you mentioned, they need to run just when there is an energy debt.

Short haul trucks are certainly more plentiful than school buses, and delivery schedules could be modified to work around this requirement. That might or might not be a simple task, I don't have any more than passing knowledge of delivery schedules.
They would obviously need to stop deliveries and store some of the excess solar for a few hours before peak demand. Maybe a 4 hr charge from 11 a.m until 3 p.m. would fit the solar oversupply and be sufficiently slow to avoid battery degeneration?

Long haul might be less suitable, their schedules might be more flexible, but battery cost might limit adoption of electric power for long haul for a while yet.

Pretty much just speculation, you might have some better insight here.

I really like the concept of the whistlecar, and remote driving, if it's reliable enough, seems like an adequate way to do it.

An advantage of remote driving which you kind of hit on is that a driver can switch from one car to another (in a very different location) quickly. So in a Halo implementation, maybe you need to move a bunch of cars, from all over the city, to the airport short-term parking to pick up passengengers coming in on a flight. One employee can drive one car to the airport, then instantly "teleport" themself to another car and drive that one to the airport. Using short-term parking, or more likely with a deal with the the airport to have a dedicated pickup location, people coming off their flight can get to the car and drive wherever they need to go. Maybe they're renting the car for a few days, or maybe they just need it to get to the hotel, and an employee can pick up the car from the hotel and place it somewhere for someone else.

Overnight, Halo can gas/charge up the cars and drop them off at the houses of people who want to use them to commute to work. One single employee can drop off dozens of cars overnight, in a way that would otherwise take two employees, twice as much time, a spare car, and twice as much driving. If the employee knows they don't need the car from 10-4, the car can be used for other things like airport runs during those hours, and be back at the office by 4, fully charged/gassed-up and ready for the commute home.

Moreover, once cars achieve level 4 in limited areas, the car can always run fully autonomously, even with passengers in it, in those areas, but is available for trips outside the geofence for ADAS driving.

Halo is only the beginning, but I think this basic model of rental cars (traditional or by-the-hour/minute) rather than taxis is the one that's going to succeed.

It certain can be useful, though it remains a stopgap until the software can do the job. Or better still, it is a gradual stopgap, in that you being with human remote driving and slowly your self-drive system can handle more and more territory and the costs of the humans goes down (or their capacity goes up.)

This has parallels to the strategy for Starship, which I was partly responsible for. Starship does not do remote driving at all, just remote strategy assist. When we first put them on the sidewalks, the autonomy was not very good, we needed a lot more remote assist. Over time this decreased and today it's rare to need it. There is not actually any need to get it to 100% autonomy, the way you need a robocar to do.

This gets you on the road, serving customers, faster and with a larger service area. You may have higher costs at first --thank you crazy valuations -- but you learn all the other very important lessons about running a real business.

But as for the value of not having to drive? I think many people value that a lot, which we debated in another thread. While a delivered rental car is great, getting to do other stuff on boring routes is also great and has a higher dollar value, if you can do it. Actually, even on non-boring routes. When I do a scenic drive it's nice to be the passenger. A narrow windy road is a different story.

To be clear I'm talking about a delivered level 4 rental robocar. The car would still be self-driving in any situation where it is capable of self-driving (unless the customer wants to drive). The car could still use humans working in a central location for solving problems. Essentially, it'd be a Waymo One vehicle that could optionally be driven by the customer if they want to go outside the geofence, and could be remotely driven by the company instead of needing to send people in vans when the car gets stuck, or to go outside the geofence if the passenger is drunk/not-licensed/willing-to-pay-more.

I think it will be a very long time before we have cars that can go everywhere people can go without even potentially needing a human driver. We can't even do that anywhere right now (hence the Waymo vans on call to take over).

Self-driving cars are not going to be something only available in select locations where application-specific new infrastructure has been built. It's too great of an opportunity for places outside of the narrow geofences. In fact, it is perhaps even more valuable outside these locations, as these are places where other forms of mass transit and rideshare are least available.

Say I want to fly to Phoenix, rent a car, and then drive to some campsite and go camping. I should be able to pick up a car at the airport and either drive it myself, or have it drive me to near the edge of the geofence and then drive it from there. Then I can return it anywhere in the geofence. Waymo isn't going to offer totally self-driving cars that can do that, for every campsite people go to, for a very very long time (if they survive that long).

You can have cars which you manually drive when you go outside the service area, or in particular onto fun roads. That depends on the cost of the system. If the system is expensive, it's better if the system takes you to the edge of the service area, and you transfer into a cheaper, manually driven car. You don't have to come to the same spot.

If the system is cheap, then of course make it dual mode.

Though many people are designing robotaxis that don't have any place to drive them from, or which would not be fun cars to drive.

I definitely think Starlink can help, but putting satellite dishes on car roofs probably isn't the implementation. Starlink is working on attaching satellite dishes to the roofs of trucks, but it's probably not going to be reliable enough for remote driving purposes. As I understand it, they require constant line-of-sight.

Using Starlink to go from one ground station to another, and then using terrestrial wireless to get from the car to the ground station, is more likely, and can be useful in more remote locations.

You could have slowly moving ground stations, or easily relocatable ground stations, in places where having a permanent ground station is more difficult. Those ground stations could be remotely driven, autonomous, or some combination of the two, but to the extent they're remotely driven they'll probably have to rely on some other ground station or some other, already-deployed, reliable high-speed wireless.

Starlink has indeed talked about entering the backhaul market. It should be possible to make a mass-produced box with a starlink terminal, wifi and a 5G micro-cell, which will cost a few thousand dollars and can be stuck on a tower, hill or even rooftop in every small town in the world that can afford it, so there is nowhere urban that your cell phone won't work, including at every signalized intersection.

However, I don't think it's out of the question that they could eventually make an antenna unit which is much smaller and can fit on a car. The harder question is whether it would always see a bird with enough reliability to do remote driving. It probably will always be able to see a bird for remote assistance as long as it sees any sky at all -- with 40,000 sats up there, it will not be long before one moves into a clear patch to allow remote assist to fix a problem with strategic instructions (including just instructions enough to get to a clearer view of the sky.)

The result is a vehicle that can resolve any problem almost everywhere in the world, even unmanned, which is a useful thing, possibly worth the cost of the starlink terminal. In cities it could rely on the cellular networks. The only places it would have to avoid would be tunnels and a few special roads with stretches with no view of the sky and no local cell signal.

I could even imagine, if purpose building a system, that you provide a longer-wave band, which does not need line of sight, said band to be used only in super rare situations by mobile devices unable to see the sky, transmitted at higher power during its rare use. Obviously this band would offer much lower total bandwidth, but that's fine for a super-rare-use backup channel. Fixed stations would not use it.

Maybe it's not true on the West Coast, but here on the East Coast we have a lot of trees.

I've heard that bad weather can also be an issue for getting good (fast, reliable) Starlink service. And that's with a receiver in a fixed location.

Add new comment