Self-Driving Vehicles/Autonomous Vehicle Technology

[crs1026]

^The interesting thing for me in the video was the computer screen showing the things (other cars, mostly) that the tesla had figured out and was working to (or, alternatively, was avoiding). Certainly quite complex.
But yeah, let’s see it handle a traffic cop, or the ever-changing pylon arrangements at, say, Six Points.

- Paul

 

[afransen]

Tesla has been making very good progress. It's a bit premature to declare victory, but what they are showing is pretty impressive (especially compared to the smoke and mirrors that Waymo is passing off as an AV).

 

[afransen]

The restriction is not simply a bit of paint. It may also be actual pavement (rural municipal roads, cottage roads, etc.). Whether the wireless data density is there may be up to data providers (if profitable) or, gasp, the public purse. The limited range of terrestrial 5G towers suggest that satellite may be more feasible. In areas where it is not realistic, will urban dwellers be no longer able to individually travel to these areas because either their vehicles will no longer work and/or they have lost the skills to do it themselves? Pastimes such as cottaging, hunting, fishing and backwoods hiking, canoeing, etc. may be in for a hit.

What % of vehicle miles travelled would be subject to these conditions? Low single digit %? Perhaps <1%? Landline telephone or grid electricity took a very long time to achieve >99% penetration (if they did). Both technologies transformed life in cities many years before they reached that kind of ubiquity in remote rural areas.

Nothing is stopping AVs from operating without continuous data connection. All the processing needs to be done on vehicle to meet acceptable latency levels anyway. Connectivity is desirable for dealing with interventions that require escalation to a human operator, but they will become exceedingly rare. Your typical trip to the cottage I don't think will be a challenge. Tesla is demonstrating a vehicle that can drive on unmarked residential streets with parked cars. I don't see how a reasonably clear gravel road could be any more challenging--it would just need to be driven at an appropriate speed.

As far as AVs driving in winter conditions, I happen to be aware of AV company Gatik buzzing around GTA to test their vehicles in winter conditions. I think driving in challenging weather conditions like heavy snow will be one of the last problems to be solved, but these are conditions human driver struggle with mightily as well.

 

[Isotack]

Next up...flying cars!

 

[rbt]

But yeah, let’s see it handle a traffic cop, or the ever-changing pylon arrangements at, say, Six Points.

Detecting and following officer hand-signals which conflict with other traffic indicators (like a no-left-turn or a red light) has got to be one of the more challenging tasks for Level 4 automation.

Heck, just spotting the difference between an officer and some random idiot waiving their arms around isn't trivial either as each city has several uniform styles.

 

[rbt]

Next up...flying cars!

Move to Alaska. 4% have a pilots license and a runways are a neighbourhood amenity. We don't all fly because it's expensive and time-consuming (maintenance, licensing, etc.)

Anchorage · Alaska

Alaska

www.google.com

Anchorage · Alaska

Alaska

www.google.com

 

[afransen]

Next up...flying cars!

I doubt we'll see vehicles that can both drive on the road and fly (just too many engineering compromises). However, small VTOL air taxi like service does seem to be coming. Lilium is just one (very well funded) example of dozens of companies working on this. The enabling technology is basically there, it's now down to the engineering work to commercialize them.

 

[afransen]

Detecting and following officer hand-signals which conflict with other traffic indicators (like a no-left-turn or a red light) has got to be one of the more challenging tasks for Level 4 automation.

Heck, just spotting the difference between an officer and some random idiot waiving their arms around isn't trivial either as each city has several uniform styles.

It is something that confuses many human drivers. Even for me at police checkpoints (RIDE) I recall their handwaving to be quite unclear as to what they wanted me to do (waving a flashlight around). I suspect part of it will be giving police better tools for such situations, and part will be vehicle to vehicle communication and orchestration to negotiate intersections. AVs will need ways to break deadlocks where two cautious AVs are deferring to each other.

 

[crs1026]

Heck, just spotting the difference between an officer and some random idiot waiving their arms around isn't trivial either as each city has several uniform styles.

Nothing is stopping AVs from operating without continuous data connection. All the processing needs to be done on vehicle to meet acceptable latency levels anyway. Connectivity is desirable for dealing with interventions that require escalation to a human operator, but they will become exceedingly rare. Your typical trip to the cottage I don't think will be a challenge. Tesla is demonstrating a vehicle that can drive on unmarked residential streets with parked cars. I don't see how a reasonably clear gravel road could be any more challenging--it would just need to be driven at an appropriate speed.

I can't wait to hear what the lawyers/ insurance companies declare is an "appropriate speed" for the more challenging conditions where people overreach today. They, rather than the engineers, will make that decision. That may come as a surprise to people who take chances today.

The issue of hand directions won't be limited to the police. Look at all the construction workers, tow truck drivers, volunteer fire fighters, and even delivery truck drivers who help navigate unplanned or anomalous situations today...in a hundred different uniforms, or not. There may be a whole new universal sign language needed. And a whole new set of authorities/offenses.

I am amazed at how the world is gravitating to satellite based connectivity for just about everything, including driving. We don't have to worry about nuclear weapons or bioweapons in the future.... our survival will hinge on whether the satellites are safe. They will be the obvious target for hostile aggression. And the ability to recover from even a single loss of a satellite will be paramount. Today, a dead satellite can't be replaced over night.

I'm sure that will all seem like childplay someday, but I can't see that coming together in this decade, or longer.

- Paul

 

[afransen]

I am amazed at how the world is gravitating to satellite based connectivity for just about everything, including driving. We don't have to worry about nuclear weapons or bioweapons in the future.... our survival will hinge on whether the satellites are safe. They will be the obvious target for hostile aggression. And the ability to recover from even a single loss of a satellite will be paramount. Today, a dead satellite can't be replaced over night.

You realize that SpaceX alone is contemplating a 40,000 satellite constellation. And they will have to launch 8k per year to evergreen them with a 5 year life. They will have spares, and the network will be tolerant to the loss of satellites (even for things like malfunction, loss of communication etc.). The bigger concern is kessler syndrome. An attack on these satellites is basically MAD as it will pollute orbit for everyone.

The issue of hand directions won't be limited to the police. Look at all the construction workers, tow truck drivers, volunteer fire fighters, and even delivery truck drivers who help navigate unplanned or anomalous situations today...in a hundred different uniforms, or not. There may be a whole new universal sign language needed. And a whole new set of authorities/offenses.

Most construction flag operators use signs, as well as the liberal application of cones. I think AVs will manage. I think you should be more worried about people maliciously putting cones in the street or tampering with signs.

 

[afransen]

Daimler has partnered with Waymo for autonomous trucks. This is likely to see autonomous trucks enter full service everywhere in North America before 2030. The unfortunate side effect will be the significant decline of rail freight. Autonomous trucks are believed to be able to transport freight at a lower unit cost than rail.

Waypoint - The official Waymo blog: Waymo and Daimler Trucks partner on the development of fully autonomous trucks

blog.waymo.com

Autonomous Trucks Could Disrupt Rail and Transform Logistics

ARK expects autonomous trucks to reduce the cost of trucking from 12 cents per ton-mile to 3 cents during the next five to ten years.

ark-invest.com

They would also need to be electric, I think, to have a chance of being cheaper than rail. And electrification may come to freight rail as well to help balance the economics. The big value prop will be the speed of long distance trucking when it can roll nearly 24h per day without needing expensive team drivers. May take a bite out of short haul air freight (for ecommerce, as an example).

 

[kEiThZ]

I am amazed at how the world is gravitating to satellite based connectivity for just about everything, including driving. We don't have to worry about nuclear weapons or bioweapons in the future.... our survival will hinge on whether the satellites are safe. They will be the obvious target for hostile aggression. And the ability to recover from even a single loss of a satellite will be paramount. Today, a dead satellite can't be replaced over night.

Hence the creation of the US Space Force, the transformation of the French Air Force to the French Aerospace Force and the standup of our own (very small) space command here in Canada. Everyone is acutely aware of how vital space is and the emerging conflict in that domain.