A new service in Los Vegas called Halo allows customers to request a rental car from them. The car is remotely driven by a human over the 5G data network and deposited for the customer to use as a rental car driven by the customer. When the customer has finished, the teleoperators bring him back to the base or to the next customer.
This achieves a form of something that I described in 2008, the “Whistlecar”. I envisioned this as an early form of robocar, for a world where it was too difficult to make a car capable of driving passengers safely on fast and complex roads, but where it was possible to make a car that delivers on its own, sticking to simpler routes. and situations, and not endangering any passenger. In this situation, it is possible to render a service quite useful. You can either see it as a carpool / rental car delivered to you and let you drop it off anywhere, or as an ‘Uber’ style service where you drive.
This is not quite right, as the driving is performed by a remote operator. It was a mistake to assume that simple unmanned robocar operations would occur before data networks good enough to be used for remote driving. In fact, it is possible that the leaders of the robocar made the whistle (so named because it was like Roy Rogers’ horse, which he could whistle, and would come to him to ride) but did not wish to. because they have many bigger ambitions with the complete robotaxi.
However, in the absence of the robotaxi, it is very useful. Existing carpooling usually requires you to travel a certain distance to get your car. The supply and location of the cars can be random in the case of a non-station carsharing like “Share Now” (which you can at least “return” anywhere in a large area if you can find parking) or it can be fixed, like in ZipCar, where you have to return the car to the place where you picked it up. Neither is very practical, but Halo fixes that problem as long as they can get you a car pretty quickly.
Alternatively, this can also be seen as a great way to use remote drivers. Have full-time remote pilots – ie. those who drive even when you are in the car – is only slightly more efficient than a regular taxi. You still have to pay a human driver to bring the car to you and drive you. However, there are these advantages:
- No human is needed while the vehicle is idle, waiting for a fare. Likewise, the cost of waiting for the car after it arrives is minimal.
- Drivers can work in one place, anywhere hundreds of kilometers away, provided they have a good network. They don’t have to drive far from home and have to go home on their own at the end, as many Uber drivers do.
- There is more privacy in the car and room to carry one more person, although there are downsides if the customer damages the car with no one to watch.
With a service like Halo, it can be a lot cheaper because the customer is driving. Although cheaper, this of course does not allow the customer to do anything else during the journey. Still, for longer trips, where the Uber price of $ 2 / mile can get very expensive, this can be a big win.
A more difficult question is whether cellular networks are really up to the task. It is an open question whether remote driving, even with a lot of bandwidth and low latency, is safe enough. Presumably cars contain additional sensors and calculations to try and protect themselves if the network connection is interrupted even for a short time, but this is not a well tested system. On the other hand, remote drivers benefit from a wide “enveloping” view which can, in theory, give them a better chance of seeing everything than a human in the car, even with mirrors. Time will tell us.
T-Mobile is an official partner here, to show off their 5G rollout in Las Vegas. 5G is often seen as a silver bullet, but it’s actually just the latest LTE combined with short-range microwave microcells for extra bandwidth and latency efforts. It is the latter that would make the difference here. The cars themselves don’t need as many modifications as a robocar, including a LIDAR, although that can help them. Remote drivers will know that cars should not be driven in areas of poor network quality, even if this involves making long journeys, to minimize breakdowns.
Many other potential benefits were included in Whistlecar’s vision. One of the most important was the ability to deliver the “right car for the trip” like a robotaxi can. Ideally, the Halo fleet would consist mostly of narrow individual cars rather than sedans and SUVs. This is because 80% of trips are made by one person across the city. If a service used such cars, it would not only be much cheaper (these cars cost much less to manufacture) but would also cause less pollution and congestion.
Halo apparently keeps the cars in a depot. ShareNow cars are parked anywhere in a city, and Zipcars are in specific special parking spaces. If the journey from the depot is long, it could degrade the quality of service. Ideally, a service like Halo would want to have citywide bundles, like Zipcar, or contract with the city to use available locations like ShareNow. It will be interesting to see where this goes.
Several companies are continuing the operation remotely. Most robotaxi companies, such as Waymo, have remote operator centers, but remote operators do not drive vehicles by remote control. Rather, they simply gave them strategic advice, like where to turn or which lane to take, leaving the robot on the ground to determine the actual direction and throttle. Remote driving and latency issues aside, there is still not enough reliable network bandwidth to depend on over a large service area, but remote counseling can be done over much more bandwidth. low.
At the same time, remote advice often involves a vehicle stopping to wait for the advice, which can mean bad road citizenship for cars. (Sidewalk delivery robots can stop much faster and can be easily moved, so they can rely more on remote advice.)
Waymo experienced an issue with its remote advisory system in Chandler, Ariz., When a remote operator gave wrong information to a vehicle, causing a comedy of errors.
Full-time remote control has some advantages over a human-driven taxi, but it still requires a paid human in the loop. Remote advice can be given halfway around the world, but the remote control should be done by someone closer and probably earning a salary similar to that of a local driver. Even so, part-time remote driving can be useful in some circumstances, including carpooling as above:
- Get a car out of a traffic jam faster than a remote board if bandwidth is available.
- Take passengers outside the robotaxi service area, both to remote areas and internal streets not managed by the software, thus greatly expanding the usable service area.
- For trucks, getting them from off ramps to depots on city streets, so the autonomous driving system only needs to handle highways.
- Driving electric cars at night from parking lots to charging stations (until they can do it themselves) so people don’t need to install or worry about charging.