The Coming Disruption of Transport

[afransen]

I am surprised that a global standard for AI-friendly road marking and wayfinding isn't already out there and being implemented, at least where new roads are being built or for the core highway network.

I'm also surprised that the various developers have set the threshold for initial rollout such that one needs an all-knowing vehicle capable of handling every last scenario in North America. That makes the design process so much more complex.

I don't think re-signing the world's streets is really necessary or likely to happen. I don't think autonomy will require handling every scenario. Waymo already uses remote operation to address edge cases where the vehicle is unsure of how to proceed. Gradually the vehicles become more capable and 1 remote operator can handle more vehicles as interventions become rarer.

I would have thought that the first step would be to exploit the "simple" applications - such as fixed routes for things like slow speed shuttle bus applications. A much more finite number of navigation challenges and road conditions, and potential to strategically target road anomalies that are the most challenging. Things like the shuttle between an airport terminal and the rental car operation, for instance.

As for intermodal, it must be much simpler to figure out the AI to allow drayage between specific key customers and the intermodal yard, than to figure out the AI for the entire TCH between container terminals on the Pacific Coast and the GTA, for instance. That's a very immediate savings without boiling the ocean. In the GTA, for instance, I bet one could pick the top 20 or 30 logistics centers and automate the drayage between these and the two (soon to be three) rail intermodal terminals. A very small set of roads to figure out and equip for smart vehicles.

The simple application is highway. The intermodal yard to the destination is the hard part--and also the part most likely to require a human driver (to disconnect the chassis etc,). TuSimple is already running unmanned trucks between their terminals in the southern US.

The airport shuttle market (and it's like) is tiny compared to the TaaS market. There are companies working on short range autonomous shuttles, but you don't hear about them. They are IMO unlikely to succeed because the resources required to solve autonomy are only going to be justified by the TaaS market size and that capability will eat all the niches like airport shuttles.

 

[DirectionNorth]

We are only 15 years post-iPhone. Think of all the business models that have been disrupted in that time and how far technology has advanced. Decades is a very long time. My personal estimate is about 10 years to see major disruption from autonomy. That is well within the planning horizon for our major infrastructure investments though. If Tesla is successful, they can deploy a pretty large number of autonomous vehicles pretty quickly. They have a track record of being able to move quickly that more bureaucratic organizations (even including Alphabet) can't match. TaaS will attrite the car fleet far faster than the replacement rate. One TaaS vehicle can do the work of several private vehicles, and if TaaS is coincident with the transition to electric, a lot of low utilization older gas cars will be displaced by a relatively small number of TaaS vehicles. And the fleet size is one thing, vs share of VMT.

I don't think TaaS will replace TTC/GO. I don't think that should be a goal, nor would it be desirable. But those organizations should include in their long range planning how they remain relevant in a transportation landscape with TaaS. They need to get better at moving people longer distances at higher average speeds, and plan for how their short distance services may need to change to remain relevant. Does TTC start operating on-demand minibuses? How can terminals be adapted to support that operating model? Does GO start operating station-to-station minibus service?

Somehow we make the same arguments but disagree completely ...

I think Musk definitely has blind spots when it comes to how this technology will affect cities. It frankly doesn't matter what proponents think, the technology will have the effects it has. I think many people will pay a premium (UberX style) for a private ride. But shared rides have the potential to be a lot cheaper and that will be relevant for people using it for every day transportation and not occasional use. I think we'll see different vehicles for the different use cases. Small 2-4 seaters the size of current sub compacts for private rides, and larger vehicles for shared rides. I think people would tend to be more comfortable sharing with more than 1 other person.

I agree that 2 hours of commuting from distant exurbs is not desirable. We already have that today and when the penalty of this is reduced with autonomy and electrification it will get much worse. We need to become comfortable with the idea that we will have the toll our highways sufficiently to contain this use case and keep traffic from becoming even more catastrophic. Those tolls will also drive people into shared vehicles, even if they have to transfer to a private vehicle when they get closer to their destination. It will be like carpool lots on steroids.

People won't pay a premium for private rides (admittedly with a driver) right now. Uber and friends seems to be eating into public transport ridership more than private vehicle ownership, at least in the US.

Ah yes, I forgot about the trillion dollar public transit industry and all the lobbying dollars it can bring to bear on policy makers, as well as public outrage about not letting transit vehicles on certain rail corridors.

That is certainly an aspiration. How do we know that is the case? If people ride the bus 7km for $3.50, and a shared AEV can carry 6 passengers and can be operated for less than $0.50/passenger km, why won't people switch? Traffic is bad? Why won't that affect the buses?

Bad faith argument.

My argument was that traffic will increase (and that there will be nowhere to put it), not that buses won't be affected by traffic.

I can see people riding AV shuttles to subway/GO stations and then take the non-congested (as in, without traffic) train into town, but the congestion that would cause would have to be regulated.

 

[lenaitch]

I am surprised that a global standard for AI-friendly road marking and wayfinding isn't already out there and being implemented, at least where new roads are being built or for the core highway network.

I'm also surprised that the various developers have set the threshold for initial rollout such that one needs an all-knowing vehicle capable of handling every last scenario in North America. That makes the design process so much more complex.

I would have thought that the first step would be to exploit the "simple" applications - such as fixed routes for things like slow speed shuttle bus applications. A much more finite number of navigation challenges and road conditions, and potential to strategically target road anomalies that are the most challenging. Things like the shuttle between an airport terminal and the rental car operation, for instance.

As for intermodal, it must be much simpler to figure out the AI to allow drayage between specific key customers and the intermodal yard, than to figure out the AI for the entire TCH between container terminals on the Pacific Coast and the GTA, for instance. That's a very immediate savings without boiling the ocean. In the GTA, for instance, I bet one could pick the top 20 or 30 logistics centers and automate the drayage between these and the two (soon to be three) rail intermodal terminals. A very small set of roads to figure out and equip for smart vehicles.

- Paul

But isn't what's already happening? They started operating on fixed routes in environments that are relatively straight and flat with benign weather, and some have graduated to more complex challenges. They have to keep investors and regulators onside. I don't follow it closely enough to know if it is the companies themselves or simply eager observers who are pushing the 'just around the corner in every situation' mantra.

On issue that, to me, still needs to be sorted out is liability. There have been discussions around who would be liable should an incident occur; the company, chip maker, code writer, etc. In another post it was mentioned that Waymo employs operators or techs to monitor their vehicles. I don't know if they have the capacity to control and intervene. No doubt these issues will be sorted out in time, but it may be a bumpy road. Developers and operators don't want to be exposed and underwriters don't want unknown risk, particularly in a land that sues over too-hot coffee.

 

[crs1026]

But isn't what's already happening? They started operating on fixed routes in environments that are relatively straight and flat with benign weather, and some have graduated to more complex challenges. They have to keep investors and regulators onside. I don't follow it closely enough to know if it is the companies themselves or simply eager observers who are pushing the 'just around the corner in every situation' mantra.

These exist but are still pretty rare. I haven’t seen a single one at a major airport eg Las Vegas, LAX or Phoenix, all of which have huge shuttle fleets to aircrew lodging, hotels, rental cars, and parking lots and are in the good weather belt. @afransen (above) seems to feel there is no money in these ”low skill” applications, but I would have thought that competing firms would want to be out there saturating this market, even if only to get brand recognition going. Plus, the ability to run multiple small shuttles instead of one big bus without doubling or tripling labour costs must create huge opportunities. The Taas companies that already run these shuttles are not lining up yet.

On issue that, to me, still needs to be sorted out is liability. There have been discussions around who would be liable should an incident occur; the company, chip maker, code writer, etc. In another post it was mentioned that Waymo employs operators or techs to monitor their vehicles. I don't know if they have the capacity to control and intervene. No doubt these issues will be sorted out in time, but it may be a bumpy road. Developers and operators don't want to be exposed and underwriters don't want unknown risk, particularly in a land that sues over too-hot coffee.

This has to be huge. One has to assume that there are activist litigators out there just itching to peel back whatever initial legal framework is put in place. One can theorise that accidents would be rarer events, but court decisions and precedents are their own kind of disruption. Customers will be looking for cheaper premiums, and manufacturers will be touting these in their marketing…. but insurers don’t have data yet to underwrite policies without a hefty markup.

- Paul

 

[afransen]

Liability is a total red herring.

 

[crs1026]

Liability is a total red herring.

Oh? In a world where many billions of dollars worth of new aircraft are grounded worldwide after only a couple of serious incidents, I would not predict that AV’s will be rolled out without regulatory hurdles.

One has to think that the two places that will have the most demanding regulatory oversight are Europe and the USA. China is a huge market, and their bureaucracy may react in any number of ways, especially if they are protecting domestic producers.

The challenge will be, unlike aircraft, once the AV’s are out there, there is little or no prospect for just grounding them. Any deficiencies that come to light (this is new technology - there will be bugs) will be patched virtually - more like phones or computers. The question will be, if a “critical upgrade” is issued, will it be mandatory to install the upgrade before the vehicle can enter the public road system. “Try again tonight” may work for mac upgrades, but…..

What would be most undesirable is a Ralph Nader 2.0 scenario where the manufacturers become aware of deficiencies but for legal/liability reasons hunker down and deny/conceal the information, until it is pried out by litigation or investigative journalism. There will have to be a regulatory regime similar to air transport, pharmacology, or nuclear power where certain types of events or discoveries must be put on record. How that will work given the amount of IP involved will be interesting.

In other words…. the rollout of AV’s will happen only as fast as the regulators can digest the technology, which may be far slower than what the engineers and programmers may have available.

Except in Florida and Texas, I’m sure.

- Paul

 

[afransen]

Oh? In a world where many billions of dollars worth of new aircraft are grounded worldwide after only a couple of serious incidents, I would not predict that AV’s will be rolled out without regulatory hurdles.

The alternative when those aircraft were grounded were different aircraft with a known higher safety performance. The alternative to AEVs will be human operated vehicles with a known worse safety performance. It was the ethical thing to do (as well as politically expedient) to ground those aircraft. 'Grounding' AEVs for not being perfect when they are much better than human driven vehicles I don't think would pass muster with the public.

In other words…. the rollout of AV’s will happen only as fast as the regulators can digest the technology, which may be far slower than what the engineers and programmers may have available.

Many more northerly climes are friendly to vehicle autonomy, including Ontario. There will be a strong economic imperative to adopt the technology. Users and industry will both be clamouring for access.

 

[kEiThZ]

Liability is a total red herring.

Just ask Boeing how that's working out for them on what amounts to a basic progressive feedback stick pusher.

 

[kEiThZ]

We are only 15 years post-iPhone. Think of all the business models that have been disrupted in that time and how far technology has advanced. Decades is a very long time. My personal estimate is about 10 years to see major disruption from autonomy. That is well within the planning horizon for our major infrastructure investments though.

First off, policy can't be made based on personal forecasts. Next, the start of disruption to mass adoption are two different things. The kind of disruption that requires a complete rethink of public transportation is way more than a decade away.

 

[DirectionNorth]

Liability is a total red herring.

If it is an obstacle to adoption (which it probably will be), then it's not a red herring.

The alternative when those aircraft were grounded were different aircraft with a known higher safety performance. The alternative to AEVs will be human operated vehicles with a known worse safety performance. It was the ethical thing to do (as well as politically expedient) to ground those aircraft. 'Grounding' AEVs for not being perfect when they are much better than human driven vehicles I don't think would pass muster with the public.

We don't know whether it will be safer.

Many more northerly climes are friendly to vehicle autonomy, including Ontario. There will be a strong economic imperative to adopt the technology. Users and industry will both be clamouring for access.

Explain this statement ... ? And which users and industry are "clamoring for access"?

 

[crs1026]

'Grounding' AEVs for not being perfect when they are much better than human driven vehicles I don't think would pass muster with the public.

I can accept that public opinion is not rational. Especially when it comes to the automobile and mobility generally, the public today has a tolerance for road risk that disregards the hard numbers.

But for AV’s - there is no data yet. And even when there is, the public’s viewpoint will not start from assuming safety. Quite the reverse…many will be intimidated by the technology, and feel safer keeping control of their vehicles.

As for believing the experts when there is data…. well, respect for science seems to be on a downward slope right now. Just wait until the few accidents that do happen are reported - no doubt with some choosing to sensationalise.

Look at how little it took to convince Congress to impose PTC on the railroads….. a relatively small number of rail accidents with a relatively small death toll. One or two “rogue monster robot trucks” plowing into a crowd or a MPV, with multiple fatalities, is all it might take to induce public skepticism or provoke grandstanding pols to defy the data in the interest of a few votes.

Many more northerly climes are friendly to vehicle autonomy, including Ontario. There will be a strong economic imperative to adopt the technology. Users and industry will both be clamouring for access.

I am sure there will be intense lobbying from industry. Some pro-corporate jurisdictions will acquiesce - hence my crack about Florida and Texas. But others may take longer. No different than multi-trailer trucks, etc.which were arguably feared as a disruptor of 20 years ago. They have influenced things, but some jurisdictions dragged their feet, and in the end the longer tandems only penetrated the industry so far.

I see a big difference between autos and freight hauling…the trucking industry will be aggressively pursuing the labour savings, and has the money to spend in Washington (and Ottawa, although Canada is the tail on the dog on this one). So yes there will be people pushing for lightning implementation, and they may get results.. Whereas - private citizens may be loath to give up their vehicles, even if the data argues for that. Urban applications will arise, certainly…. but the rural user will have to see a price benefit to give up their pickup truck. (Side thought: If anyone wants to stir the pot, just start a credible rumour that government intends to mandate AV for all pickup trucks ;-) ) AV for private vehicles is a lifestyle disruptor rather than a dollar lever, and public moods can be fickle. Everyone has a phone these days, but only a few actually camp out to get the new model on the release date.

- Paul

 

[kEiThZ]

We don't know whether it will be safer.

He is going off Tesla data (which Tony Seba also uses, in the lead presentation I posted).

Tesla argues that autopilot has fewer incidents per thousand kms than human drivers and that this means Level 5 autonomy will guaranteedly be safer. But I think this is a bit of a flawed comparison. In reality, Teslas aren't fully automated. So it's closer to an airplane where pilots set and monitor the autopilot, and intervene when adverse situations develop. Tesla never discusses how many times per thousand kms, the driver had to intervene and how dangerous those situations might have been. The fairest conclusion from Tesla's dataset would be that all drivers could benefit from some automation.

 

[afransen]

If it is an obstacle to adoption (which it probably will be), then it's not a red herring.

If it is safer, it will be lower cost. Unless courts start granting damages 10x or 100x what they do in human-caused accidents. Fleet operators will self insure. There will be contracts between the vehicle/system providers and fleet operators or vertical integration that lays out how liability is assigned.

We don't know whether it will be safer.

It will be, or it won't be deployed.

Explain this statement ... ? And which users and industry are "clamoring for access"?

Will be, not are. You changed the verb tense. I am familiar with the industry. If it lowers cost, industry will lobby for it vociferously, just like it lobbies for 60 ft trailers, LCVs, etc.

 

[afransen]

but the rural user

... doesn't matter. Whatabout some guy on a ranch with a pickup truck--it's not an argument against the fact that autonomy will disrupt transportation.

 

[C_Johnson_1995]

Cruise CEO Kyle Vogt did an interview. This reddit post gives a summary and includes the link to the interview:

https://www.reddit.com/r/SelfDrivingCars/comments/v5mq6k

One of the big takeaways is that expanding into other cities isn't a significant technological challenge, but rather a challenge from the service and infrastructure side.

This supports the view that technology for autonomous vehicles is a solved problem which is demonstrated to work reliably, as seen with demo service in San Francisco and Pheonix.

Also, this means that we can't just ignore autonomous vehicles anymore. They will disrupt all forms of transport and we need to prepared. Most members of this forum will have likely either ridden in or at least seen a fully autonomous by 2030.

To elaborate on my previous point relating to the freight railways, as we will probably see the rollout of autonomous and electeic truck in the next few years, it will be important to manage their decline and try to stabilize them to ensure their continued existence in the future as autonomous trucks will result in severe losses in volumes for the railways.

Full nationalization and consolidation would be a good way to accomplish this as duplicate track and excess capacity would be able to be cut. For example, traffic through northern Ontario from Sudbury to Winnipeg could be consolidated on CP, with CN's line abandoned should traffic drop low enough.

I also think that Transport Canada should be carrying out detailed studies on how autonomous vehicles will shift the competitive balance between modes of transport and the economy as a whole as this would be very useful in future planning, especially urban planning.