Ethics professors solve the "Trolley Problem" by debating switching tracks to kill 1 person vs. 5. Engineers solve it by fixing the brakes.

Topic: 
Tags: 

Since the famous Trolley Problem has come up again recently thanks to the MIT Moral Machine, it's time for what seems to be an annual debunking of the notion.

This time, to illustrate the pithy headline above, I tell the story of why the hypothetical situation is even rarer than people imagine because of the way braking and steering systems are designed on robocars, and how their driving patterns will be designed to minimize risk.

Read all about it in my new Forbes.com article Robocar engineers prefer to solve the runaway trolley problem by fixing the brakes on the trolley

Comments

I thought of the trolley problem when I read the Reddit post about the guy who swerved in the snow in his Tesla to avoid a collision. The guy says, "I expected to hit the van so I would miss the stopped car." In the video the Tesla seems to prevent the guy from swerving into the van.

My thought was that the guy tried to treat this like a trolley problem and the Tesla said "I'm sorry Dave, I'm afraid I can't do that." I also pondered what the collision avoidance system would have done if instead one collision or the other had been inevitable.

Note that some people have been mischaracterizing what happened and saying that the Tesla swerved. That doesn't seem to be true, but from the description and the video it does seem like the collision avoidance system did prevent the car from sideswiping the van (like the driver intended).

"But what if the brakes fail?!?!"
"Add a second set of brakes."
"BUT WHAT IF THAT SET OF BRAKES FAILS?!?!?!?!?!?!?!?!?!"
"Require that if the primary brakes fail then the car immediately moves to the side of the road and bricks the drivetrain until a service truck arrives."
"So, like, you're just gonna make me STOP ON THE HIGHWAY?!"
"Are you suggesting that if you were in a manual-drive car and the brakes failed you'd keep going?"

I am hopeful that as a robocar fleet is optimized, the bulk of vehicles will be city cars with limited performance, and hence lower mass.
Collisions with pedestrians will become less likely to result in fatalities. Reducing mass though is de-incentivized because of all the other high mass vehicles on the road.
Another aspect that will determine robocar mass is the optimal range of a city robocar as battery mass is such a large proportion of all electric vehicles. It could be that having shorter range (cheaper) vehicles only requires a few more vehicles to service the customer pool. If so the vehicles may eventually be very light.

I'd be interested in the numbers on that one. A lower range vehicle is less expensive, but if you wind using supercharging during the day because of the lower range, are the costs going to outweigh the benefits?

For commuter vehicles, they want enough battery to handle the morning rush, recharging in the mid-day to handle the evening one. If they don't have enough for the rush they go on a lower usage duty cycle. Not the end of the world, I guess. It means they take longer to wear out, and they are more urgent in charging need.

Electricity from 2pm to 9pm is much more expensive. Most cars will want to avoid charging then, and thus to have batteries that can handle their work. On the other hand, you can have the batteries at charging stations, not in the cars, and then swap or just transfer-charge (at a loss.) If the losses of transfer charging are less than the cost of carrying the weight, it can be a win. Since, as you say, one of the costs of the weight is extra accident risk, that would need to be factor in.

However, since solar is on a tear, dropping in price, we may see that the peak for charging changes. The solar panels will spit out their power from 7am to 5pm, best at noon. However, if you aim them southwest instead of south, you can move their peak later in the day at cost in overall output. While solar is variable (which is a downside) it turns out that for car charging, solar positioned for peak robocar charging demand might be very efficient. "It's sunny -- hey robocars, come here and charge with me!"

Will commuter vehicles really need to recharge during the day? How many miles of range do you think they'll have? It'd be nice if you could exclusively charge them overnight. Unless the peak for charging changes.

This is the tradeoff. You can either elect to have a bigger battery, good for most of a day's service, or you can have one perhaps 1/3rd to 1/2 that size, good enough for "one rush" with 9:30am recharge, possibly more service and 1pm recharge. 1pm recharge might need fast-charge, which has its own costs.

So we need to make a spreadsheet, which I have not yet done, based on:

  • The extra cost of the larger battery (mitigated by its longer life as each cell is doing one cycle per day instead of 2-3.) The cost of capital tied up in the battery.
  • The cost of the weight and size of the larger battery in vehicle design and energy usage.
  • The higher cost of recharge during the day, including travel to the charging station and downtime due to charging

I don't have those numbers so I haven't made the sheet, but I understand the intuition that the larger battery is probably the win.

However, the full day battery probably needs 200 miles of range if urban, possibly a bit more suburban. And you only want to use from 20% to 80% so really you need 330 miles of range. Which is an expensive battery today. Though it lasts twice as long.

The other option is battery swap. With battery swap you have the cost of swap, but you eliminate the size and weight cost of the battery in the vehicle. You charge all batteries at night. Battery swap doesn't work for humans, it's fine for robots.

Another issue may be space at recharging stations, which might make it worth doing a bit of charging in late morning. Particularly if there is a "solar peak" as solar grows, it may make 10am-2pm become one of the cheaper electricity prices of the day. Though in summer this could also be the time to chill down water tanks for use in air conditioning from 2pm to 6pm.

Another option is to have a lot more vehicles. 200 miles a day is a lot. If you could cut that to 100 miles a day, and have twice as many vehicles (actually less than twice as many vehicles, because you'll have less need to drive around without any passengers), would that make sense? In places where parking is cheap, I think it might.

It depends where you serve. If you're driving highways at 60mph, with two rush hours from 6am to 9am and 3pm to 7pm plus some fringe and the noon mini-rush, pretty easy to rack on 200 miles. If you stick to the city at 12mph average speed (though I hope that improves) then you might do only 100 miles/day. But average NYC taxi drives 170 miles/day from what I recall, though that includes more empty miles than robotaxis will have.

Sure, it's easy to rack on 200 miles, but you can also choose not to. Instead of having one car handle 4 solo passengers traveling 20 miles each between 6am and 9am, you can have two cars that handle 2 passengers each during this time. The cost of that is, as you point out earlier, capital and parking. There are a number of benefits (including the ability to have smaller batteries), but one big one is that you'll likely have a lot fewer dead miles. Rush hour commutes tend to be heavily skewed in one direction. To handle 4 solo passengers between 6am and 9am is going to require a lot of miles driving in the opposite direction of the demand, and while you'll be able to fill that with paying customers in some cases, you won't be able to fill it with paying customers for all of the miles in all of your cars.

If parking is cheap (especially as long as interest rates remain low), I think this will be the way to go, in some areas.

I'm a big fan of pooling once you make it frictionless, but I am presuming you are still going to have some solo commuters. It could be there will be more emphasis on pooling the much longer commutes. To answer these questions requires some heavy duty activity simulation. The data are out there in the NHTS but it's no minor project.

There is reverse commuting, and in fact there will be a surplus of supply and road capacity for it, so those folks may get more solo rides, because on the vehicles doing the longer commutes, you don't get in two many round trips in rush hour.

Of course today, "rush hour" is no round trips during rush hour at all. All the cars simple come in from the burbs and park all day. The robotaxi fleet could be big enough to do that, but that's wasteful so you will try to get in some reverse trips (ideally paid.) But if there's not enough demand for that, then you either have the larger fleet, or you price to encourage more pooling. It is one of the issues I need to resolve with commuter vans. The most efficient pooling is in vans, but the vans have much less demand during the day and in the reverse commute direction. Not zero demand, and you can still serve just a few people in a van, but you want to avoid that. So there's a big model to be built to figure out the optimum distribution of vehicle sizes and trips.

For simplicity I was ignoring pooling. Pooling is to a large extent a separate consideration, as there are a lot of people who simply wouldn't do it, even if it were free. For a 20 mile commute (40 miles round trip), 5 days a week, 52 weeks a year, with a $0.25 incremental cost of driving, commuting costs are $2600/year. That's significant, but it's hard to see how pooling could possibly so seamless that it's not worth $2600/year to avoid it. (My own commute is only 5 miles and under 15 minutes. And my incremental cost of driving my subcompact hybrid is probably closer to $0.15/mile. It's even less likely you'd get me to switch to pooling to save $390/year.) Some people will pool, and robotaxi services need to have pool cars available for those people, but there will also be demand for solo commuting, if the price is reasonable. (I'd probably pay $0.30/mile for a robotaxi to work if it could reliably arrive within 2-5 minutes of when I call it. More, though not much more, if the service were reliable enough that I could get rid of my car altogether. Possibly not if I could buy my own personal robocar for a similar price, though.)

Anyway, back to ignoring pooling (because that's really a separate market). On the one end of the spectrum, you could minimize your number of cars. On the other end of the spectrum, you could give every single commuter his own dedicated robotaxi. Either of these extremes would be wasteful. The former wastes energy (as it requires a lot of dead miles). The latter wastes capital and parking space. Somewhere in between is best, but I think that for most commuting purposes, the most cost effective solution is much closer to the latter than the former. No vehicle would be *dedicated* to any particular person, unless that person paid much more, but if you have 100 people working at an office park, it might make sense to have 50 to 75 commuter vehicles parked there or relatively close to there from 9 to 11. During the lunchtime rush (which might very well consist of delivering food as much as carrying passengers), you'll probably rotate a lot of them out, but as the lunchtime rush dies down you might get back up to 50-75. All day you'll rotate them out for demand nearby. Maybe the airport is nearby, and you might send some vehicles over there to replace the ones picking people up at the airport. Some people are going to be starting their commute between 9 and 11, and you can rotate some vehicles out to handle them (or to fill in other locations where there was a shortage caused by vehicles going out to handle them). By 4:00-4:30 you probably want to start sending vehicles to the office parks from nearby locations (and filling in the gaps from farther away locations) in order to get ready for the rush home.

This seems like the best way to handle things where I live and work. Where I work, the rush hour is fairly tight. If I arrive at 8:15, maybe 10% of parking spaces are full. At 8:30 it's maybe 20%. At 9:05 it's 90%. You're not going to get much reuse during the morning commute here.

I think we will see a certain amount of frictionless pooling, as I describe. The more frictionless it is, the more you would do it to save $2600.

Except it's not $2600. As I wrote last week, your commute is definitely not "incremental miles." You bought your car probably with commuting as the number one purpose. If you are not going to commute in it, you will seriously think about not owning at all. When you buy your car, you have in mind a certain amount of travel in it. You must consider that travel as costing the full freight, ie. taking a share of the monthly non-use costs. It may cost $200/month for your car to sit around, but the reason it is sitting around is to meet the needs you bought it specifically for. You can consider as incremental trips you did not plan to take.

It's also possible (though speculative) that cities will solve congestion, once they have the opportunity to, by congestion charging/metering of some sort. If they do so, then the savings from pooling will jump up. In fact, they will jump up exactly enough to get the excess solo cars off the road, because that will be the rule. By definition the savings from pooling will be enough to convince the overdemand to pool. If the roads have capacity for 100,000 trips and 140,000 people want to travel, you raise the price to whatever it takes to make 60,000 to 80,000 of them want to pool, untill there are only 100,000 vehicles using the capacity. Or you force them in some other way.

We've never done this for two reasons -- we didn't have the technology, and it takes some political change too. I see that coming.

Commuting was not the number one purpose for which I bought my car. If I was not going to commute in it, I wouldn't seriously think about not owning at all. Commuting is definitely "incremental miles" for me.

I don't think pooling could ever be frictionless enough that I would commute to work that way. Not for my current 5 mile, less than 15 minute, commute. I'd have to be the last one picked up and the first one dropped off, door-to-door, and I'd have to be able to be picked up in 2 minutes or less from the time I hit the "pick me up" button. It'd be impossible. (I stayed 20 minutes later than usual at work last Wednesday. I didn't know I was going to need to do this until I did it. I didn't know when I was going to be done until I was done.)

For other people, with longer commutes, or with less efficient cars, or with a tighter budget, or just with different preferences, sure. But I can't see myself choosing pooling so long as I have the commute I currently have.

For people who work in the city, pooling might very well be a good choice. For many people who work in the city, pooling already is a good choice, and it'll only get better. But for people with short suburban commutes, nah. It's already cheap enough to commute solo.

If you have not read my transit plan, you might use a phrase like "last one to be picked up." In that plan there is no last person to be picked up or first to dropped off. It is not as perfect as a private ride, but it's only 2-3 minutes longer and perhaps 1/4 the price for many commuters -- which I think would make a difference.

And yes, you would be picked up shortly after signaling -- of course. People would not accept it as competition for their private car if it did not pick you up quickly and get you there in only slightly more time, after all.

Well, they would if it actually were faster than your private car, which it could be if pools get preference on the road. As they do today with carpool lanes. And that's why people line up to get picked up by random strangers to carpool over the Bay Bridge or down I66 in Washington. If the vans can use the old subway tunnels after the tracks got taken out, they are going to do better than cars on the surface.

Good point. I admit I TLDRed some parts of your transit plan, but now that you point that out (and I think I can guess what you mean by it), I'm going to go back and read it more carefully. (Edit: Nope. I guessed wrong. Transitions don't work when you have a 5 mile commute. Though if you've written about a way to do transitions from one vehicle to another while both vehicles are moving, that would be fun.)

Still, for me, with a commute of only about 12 minutes, 2-3 minutes extra is too much. Even if it's free. 2-3 minutes of my working time costs more than my whole commute costs. And with a 12 minute commute I'm not going to get work done on the commute. Paying bills, doing online shopping, maybe. But not the real work that I'm commuting to get paid for.

It sounds like that 2-3 minutes doesn't include the time between pushing the button and getting picked up. So that's even more unacceptable, for me.

Pools aren't going to get preference on the road on my commute. And they shouldn't get preference on the road anywhere, in my opinion.

Pooling will work great for lots of people. People with longer commutes will save a lot. People who work in places where there's a high volume exodus and not a lot of loading/unloading space might even be able to board much quicker by pooling.

I don't think it'll work for the average commuter, though.

I wrote about that for trains 15 years ago. For now I have not considered it workable yet, though it may come time to think about it. Obviously anybody who takes transit today, which can double or triple trip times, is not concerned about 2-3 minutes on a 12 minute commute. So there are lots of people who would be very happy to accept that if it made it 1/3rd the price.

More to the point, the whole reason for this is that in some places, there just isn't enough road capacity for everybody to commute in a private vehicle with 1-2 people. We get traffic jams because of this. So some form of group transport is necessary for some segment of the people. This is the best form of group transport I have yet to imagine.

Whether people who don't like even a short extra delay would be convinced to take it depends on what the governments want to do in order to push enough people into group transit. It is actually in nobody's interest, including the super time conscious commuter, to have more vehicles enter the road than it had capacity for. So you want to limit traffic to match capacity.

You can do that by charging money. You can do that by offering incentives beyond money to those who pool. But somehow a fraction must be convinced. One of those incentives is time through access to dedicated right of way like carpool lanes.

Would you not take the pool if it took a guaranteed 15 minutes and your private commute would take from 12 to 25 minutes depending on traffic?

Yes, I would take the pool if it took a guaranteed 15 minutes and my private commute would take from 12 to 25 minutes depending on traffic. If the government decides to create a problem where none currently exists, they might be able to force me to work around it. And they'll no doubt create lots of unintended consequences in the process.

I would support a road usage fee based on the size of the vehicle and the time it's on the roadway, if, and only if, it replaced gas taxes and raised a similar amount of revenue. I wouldn't support much more than that. And I'll gladly pay the fee for driving my subcompact vehicle less than 30 minutes to and from work every day. It might come out a little bit more expensive than gas taxes, since my vehicle is so efficient, but it'd still only be pennies a day.

The details of your particular commute are not the concern here. One of pooling's main values is reducing congestion. If somebody's commute has no congestion, then pooling is not a big issue for the government, though it does save energy.

I am presuming if pooling is under consideration that the commute is unpredictable due to congestion, and may take 12 minutes with no traffic but up to 30 when it's bad. That's certainly how it is around here, and those are the numbers for a commute I used to do.

Which is why governments, to encourage pooling, make carpool lanes. They don't do a great job of it (carpool lanes don't work but HOT lanes do) but the general idea makes sense, to encourage pooling with better and more reliable travel times.

HOT lanes and carpool lanes both make traffic problems worse. Carpool lanes take traffic lanes away from non-carpool traffic and reduce the disincentive for people using them to find a different route, to avoid commuting altogether, or to commute at off-peak times. HOT lanes are not as bad as carpool lanes in taking traffic lanes away from "regular" lanes, but they're much worse in reducing the disincentive for people using them to find a different route, to avoid commuting altogether, or to commute at off-peak times.

I do agree that there's a tragedy of the commons problem if you don't charge for road usage, but I would argue that this problem is best solved by charging a flat fee to everyone based on the amount of space on the road you're using multiplied by the amount of time you're using it. Pooled traffic would pay less under these scenarios, but only less in proportion to the reduced amount of congestion that they're causing.

Ideally, the land for the roadways would be bought by those voluntarily selling it, and the congestion fees would pay for the purchase of that land (along with maintenance costs, though it might make sense to charge for maintenance differently due to the 4th power law; pooled traffic would pay more for maintenance costs). I wouldn't get my hopes up about that one, though.

(Once robocars become fairly common, it will probably make sense to have lanes/roadways reserved for robocars only. These lanes would be faster. But the purpose behind that would not be to mess with free market incentives; it would be to reap the significant benefits that can be achieved on roads that robocars don't have to share with human-driven cars. A properly-built robocar highway system would essentially never have traffic jams except when there's a crash. The wait time would be kept to the interfaces between the robocar system and the general traffic system -- off-ramps leading to waiting areas out of the way of the through-traffic. Interestingly, this might be feasible to do privately, especially in areas that aren't too heavily built-up yet. Or by going underground - literally - a la Elon Musk's Boring Company.)

Do you have a research cite on the HOT lane disincentives? A proper HOT lane strategy is to know the volume of carpools, then set a price to bring in just enough solo drivers to make best use of the lane capacity, but not overfill it. You are saying that now people who would have taken the train drive because they can buy HOT access? There really aren't that many people who take the train compared to use the road in most cities, so I would be surprised to see that.

Part of my plans are carpool lanes which provide access only for true induced carpools. ie. no family pooling or other people who would have ridden together naturally. Induced carpools are only 10% of carpools, which is why carpool lanes fail. I will be outlining ways to enforce that in a future post.

The 4th power law does make it interesting. A back of envelope calculation does show the people in a small van doing a lot more road damage per person than a solo rider -- perhaps even 50x as much. What I don't know is what the cost is. If one unit is .01 cents/mile then .5 cents/mile is tolerable. If one unit is 1 cent/mile and the van does 50 cents/mile per person, that dwarfs the other driving costs.

That needs some more in depth research. The 4th power is also just an approximation so one would want to learn the values for the real weights. My guess is a 15 person vehicle might weigh 5 times as much as a one person vehicle, plus the weight of people in each case, but that's just a guess.

Update. Some quick looking found estimates of road damage from 4 to 15 cents per ESAL, where an ESAL is the damage from a truck with 18,000lbs per axle going over it. A regular car 4,000lb car is .0003 ESALs and thus a 1,000lb solo car 1/256th as much, and a 6,000lb van 5 times more. But that makes the solo car van about 0.0000012 cents per mile per damage, and the van .0006 cents/mile/person -- so 50 times as much, but 50 times such a small number that it doesn't matter.

I concede the 4th power law point.

The concept of "induced demand" has been well studied. I don't know if it was specifically studied with regard to HOT lanes, but over the long run, increasing highway capacity does not decrease congestion. In most areas trains are not a viable alternative, but moving closer to work might be, or working closer to where you live, or working from home a few days a week, etc.

That argument goes for inducing carpooling too. Carpooling increases highway capacity, but in the long run that just provides incentives for more people to use that road. A HOT lane that exempts carpoolers from tolls will quickly fill up with carpoolers once carpooling becomes feasible. What do you do then? Start charging carpoolers too?

Limiting the number of cars on a road to maximize throughput is great, but HOT lanes are not the only way to do that. You could just as easily limit access to a roadway in other ways. Even more easily with robocars, as you can much more strictly enforce variable speed limits based on traffic conditions on them. You could do that with or without giving priority access to rich people, but it is a mistake to treat carpoolers drastically differently to try to induce carpooling.

Carpooling is not a long term solution to traffic problems. And while carpooling is a more efficient use of the roadway, it's not so much more efficient that we should charge non-carpoolers such a drastic premium for using the road.

Yes, I am quite familiar with this, and so my long term plans deal with it. However, trying to get people to understand the whole path takes time, and carpool lanes are a well understood metaphor for offering a faster trip to people if they carpool. The long term answer is metering. If you don't let more cars into a lane than it can handle, you will eventually stop inducing demand because you turn away the new demand rather than let it add to the load. In one extreme example, you could say, "Sorry, no use of this freeway if your house was built after 2019." You would not actually do it that way of course. The best way is money, raising the price until demand <= capacity, but that's politically difficult.

That sounds more reasonable. A van with 6 people in it should pay a higher toll than a compact car with 1 person (or 0 people) in it, not a zero toll. I'd be hesitant to raise the toll too high rather than turn people away, though. It's a complicated situation because many of the drivers are taxpayers who paid a lot in taxes to get these roads built. Now you're asking them to pay again. Prices that are too high indicate that you haven't expanded capacity enough, and that's only possible if you have a monopoly on the road system. In most situations there will be alternate routes, so simply closing the road to new traffic (or increasing the time between green lights on the ramp meter or modern equivalent) is a reasonable solution.

Carpooling shouldn't be artificially induced. When governments try to induce certain behaviors, there are always negative unintended consequences. The long term answer is getting the government out of the road business.

Fully private roads do exist, but are and will be extremely rare in the USA. Routes compete but only to a limited degree, usually some routes are inherently superior to others by being shorter.

But, politically, road charging is hard in many countries. I am surprised that egalitarian Sweden was one of the few pioneers.

But whether you are a government or a private road owner, the right approach is to not allocate more usage to a road than it has capacity for. In the coming world where "a smartphone in every car" is possible, there are now many options to do that, not just road use charges.

But either way, to take cars off the road, there will be pooling. Yes, there are other options -- trains, telecommute, time-shifting -- but there will be pooling induced for those for whom those options are inferior.

I'm sure there will be pooling, and I'm sure it will be more than it is today. Hopefully it is induced by the natural savings that come from it, and not by artificial government incentives.

Interestingly, if governments do provide too many artificial incentives to pool, robotaxi companies might very well charge more for a pooled vehicle than for a solo one.

The problem is that to induce enough pooling from just natural savings may require setting the price too high; a price that governments would not want to set. If they add a $20 usage fee that would push some to pool to save on it, but would leave those without a good pooling option paying a charge the public would find untenable. "Pool and you get there faster" will be an easier sell. The roads are, for better or worse, owned by governments, who are responsible to the general population of voters.

It's not my goal to induce pooling. That's actually the mistake that the people who make government rules make. They confuse one of the means (pooling) with the goal (making commutes more pleasing). In the extreme, for some of them pooling becomes the goal, with no regard for the actual human beings that we were originally told that pooling was supposed to benefit.

A minivan that holds six people causes the same amount of congestion whether there are six people in it or one. It should pay the same congestion fee, though obviously that fee would be split six ways or be covered by only one person.

If that doesn't "induce enough pooling," then pooling isn't a painless enough solution to the problem of congestion.

"Pool and you get there faster" will have unintended consequences. Yes, you'll induce more pooling, but you'll also induce people to cause other problems while achieving the pooling. Are you going to have HOT lanes literally everywhere? Probably not, in which case one of the unintended consequences of "pool and you get there faster" will be to greatly increase traffic on local roads.

Let's say the HOT toll tends to go to $20 during the morning commute, and that fee is waived for cars with X or more people in it. That means certain people are willing to pay $20 for the faster travel. A robotaxi company (ignoring your advice) could charge $15 to someone who is picked up last and dropped off first. On the one hand, that achieves the artificial goal of inducing pooling. On the other hand, it causes more congestion on local roads and increases the travel time of those who can't afford the extra $15. In the short run it causes more people to commute longer distances during peak hours, as their commutes are now subsidized by $15/trip. In fact, it makes it cheaper to commute during peak hours than it is to commute during off-peak hours, so long as you're willing to pick up an extra passenger along the way. And in the long run, due to induced demand, it doesn't even solve the problem of congestion on the highway, as someone new will come along to fill the spaces on the road abandoned by those who were artificially induced to carpool.

Well, my new Drupal theme is not handling deep indent well. Sigh.

The reason that you offer the poolers faster times is that the way lane management works is you have to give it to somebody. So who do you give it to, if politics forbid you from giving it to the highest bidder?

Because you can't just make one car get a better trip, you have to make at least a whole lane full of cars get one, or ideally a whole highway.

And why not congestion charging/lane management on all congested lanes, during rush hour? The error was treating the roads like a commons, and thus they get overused and congested and less valuable to all.

If politics forbid you from giving the lane to the highest bidder, you could give it to everyone equally. Ramp metering can be very effective at reducing congestion. Congestion-related variable speed limits (whole road or, ideally, per-lane) can also help in certain situations where ramp metering doesn't, though they are hard to enforce (but will be much easier to enforce against robocars). But to a large extent I think the answer is that politics shouldn't forbid you from charging people to use the roads. My biggest problem with HOT lanes on interstate highways isn't the toll part; it's exempting carpoolers from the tolls. (Though the toll part can become a symptom of the problem if the government is abusing its monopoly over the roads.)

Why not congestion charging/lane management on all congested lanes, during rush hour? On limited-access highways, sure, that'll work. For local roads, I'm less sure how you'd implement that, especially in a way that is politically feasible. But I'd be open to ideas, as long as they don't give special perks to carpoolers beyond the natural benefit of being able to split the natural costs of travel. As I said before, in a slightly different way, an equal-sized van should cost the same in congestion fees whether it's carrying six people or two people or one person or no people. It causes the same amount of congestion in any of those cases.

If you're talking about replacing all of my miles, then it's total cost that I'm comparing against. If you're talking about replacing only some of my miles, then it's incremental cost that I'm comparing against.

An important thing to know, when figuring out the cost to replace all of my miles, is that most of my driving is not solo driving. Most of my driving is with my kids in the car.

Another important thing to know is that my car is already several years old. My total cost of ownership starting right now is fairly close to my incremental cost. If I sold my car right now I'd only get about $6,000-7,000. If I sold it in 3 years after driving it 36,000 more miles (which is probably about what I'll do) I might get $2,000-3,000. So my total cost of ownership over the next 3 years is only $0.11/mile more than my incremental cost of ownership.

While you may not have purchased your car with commuting as one of the goals it was meant to do, many people do. I am very surprised if it's the case that, while commuting was not the number one purpose, it was not still a key purpose. But for some people I am sure that is the case.

The trick is, it just has to be a factor. We can group people into a few camps.

  • If I did not have to commute, I would not buy the car, I would rely on alternatives like robotaxi
  • I would still buy the car even if I never commuted in it
  • I am somewhere in the middle, and the lack of commuting will influence, but not fully control my decision on buying a car.

I suspect most people are in that last camp. And as such, some would buy and some not.

Put another way, if you ask the car buyer, "So why is it worth paying $200/month to have the car sit there even if you don't drive it and miles cost 25 cents when you do?"

If they never think "handy for commute" only then can you call the commute miles incremental.

I don't think there's a single right answer for how to measure the cost of miles. When comparing two different scenarios, and in both scenarios you own the same car, but in one scenario you drive the car more, the right measure is incremental miles.

For me, I don't think I would get rid of my car completely unless there was a convenient and affordable way for me to travel long (50-100 miles) distances door-to-door with my kids. That, for me, is the killer app that would make me consider going carless.

Battery swap is interesting because it does appear to offer the advantage of using more but (much?) cheaper batteries.
Longer battery lifetimes would appear possible due to reduction is supercharging frequency. In addition batteries could be optimized for high capacity and slow charging, rather than trying to be all things to everyone, possibly further reducing battery cost/complexity.
Obvious downsides include the requirement for standardization, and the (hopefully simple) infrastructure. Possibly the requirement to pick up a physical battery reduces the number of places a vehicle can go to recharge or spend the night.

I have started a new thread since my theme is indenting too much, need to keep hunting.

Governments focus on one goal in their lane management -- reducing cars. Revenue (tolls) are secondary. Any step they take has to improve the roads for everybody, not just toll payers, so that's why the HOT lane is free to carpoolers. The goal is not to sell a better ride, it's to get cars off the road and give everybody a better ride, and make the road more efficient in terms of dollars spent per person moved.

Carpool lanes fail at that because they don't induce enough carpools. Only 10% of the pools are induced. And they often don't fill the lane, meaning they actually make congestion worse in the non-carpool lanes, failing at the social goal.

So no, they are not going to charge the carpoolers. The carpoolers are the entire point of the lane. The toll-payers are just there to move some rich folk's cars out of the regular lanes into the otherwise wasted capacity in the HOV lane.

Better, but more futuristic, is to manage the whole road, not just one lane. Then take various steps to meter it so that you never put more cars on it or any lane than it can handle. Metering lights can't do that because you get backups at the lights if you try to meter that much. (Current metering lights just slow traffic when congestion is getting too high. Once they do their job and reduce congestion and flow starts again, they drain out those waiting.)

I believe full metering has to be done at home. ie. your phone tells you when you leave. So you wait at home, not in a metering lane.

The government's goal is to move as many people for the least expenditure of public money. Or, since roads already exist, to maximize the capacity they have.

If they don't, they need to build other capacity, which is hard.

Reducing cars is not a good goal, but even if that were a proper goal, making the HOT lane free to carpoolers does not do that. It in fact does the opposite of that. Inducing carpooling only reduces cars in a completely closed system, where no one moves, where no one changes their place of employment, where no one changes their route, where no one changes to or from working at home, where no cars are ever used to carry people from their homes to the pooled vehicle, etc. If the goal were to reduce cars, turning the HOT lane into a jammed up general traffic lane, or removing it altogether, would be the way to go. But of course that isn't the goal, as you allude to later.

You're right that ramp meters can cause backups at the lights if you try to meter too much. You also suggest a possible fix to that when the roads are used by robocars (meter at the home). Another possibility is to slow down traffic on the main line (with variable speed limits) prior to the merge. This could be done in all lanes equally, or traffic could be slowed down more in the lanes that are actually involved in the merge. As I said in my first message, though, enforcement of variable speed limits is tough. And if the goal really is to reduce cars, causing backups at the lights is actually a good way to do that. People will get frustrated with the backups, and will find new routes, will adjust their schedules, or will move somewhere else.

Your last two paragraphs suggest that the goal is actually something other than reducing cars. I'm a little unclear on what exactly the goal is, but it seems that raising revenue can help with it, both by reducing the amount of public money that has to be spent and by raising money that can be used to build other capacity. This is most true with high-speed interstate highways, where the people who use the roads are most disconnected with the people who pay for the roads. It's least true with local roads, where the benefit of the road goes mainly to the homeowners and business owners who live there. The tensions between local governments and state governments and federal governments are another reason why problems and inefficiencies happen.

The government's general goal is to allow all people to do their commutes (ideally the way they want, but of course how they want depends on what's available.)

So it can build roads, or transit lines. It seems to love if it can get people to transit lines, and indeed, if they could get high utilization, that would be the cheapest way to serve the need. But people don't want to use transit. It's inconvenient and slow unless it gets a private ROW and often even then. So a lot of people want roads.

So it builds the roads it can with the budget it has. Now it wants to move as many people as it can with those roads, otherwise it's going to need to build more transport infrastructure.

The best way to move more people on those roads is to get them to group together, presuming they all want to go at the same time. You can also convince them to switch times but that's not easy. And yes, you can get them to move, which they will if the commute is much better elsewhere, but that's a failure.

So if I have a road with capacity for 2,000 cars and I have 4,000 people to move, you can do the following:

  • Let all 4,000 people try in their private cars, and have the road congest, ruining it for all until people move away.
  • Increase average occupancy to >2 per car, and have demand == capacity.

How to increase the occupancy? Well, you can try fees per car. But those fees probably have to be very high. If you charge $20 per car per day, that's $4,000 per year, which a lot of the population can't afford. So they can pool 4 to a car, and now it's $1,000 per person but that's still a large burden for many people. But on the days they can't pool (and there will be those days) now that have that price.

I don't know what price you need to raise the occupancy to >2 on average, but I suspect it's pretty high, higher than needed. It probably more than pays for the road, actually. But people don't get that their taxes got lowered because road budgets were spent long ago.

Making the carpool free isn't the big deal. It's the faster lane that's the big deal. Though of course, perversely, if a carpool lane works really well then there is no congestion in any lane, and now the money makes the difference. It's a balance and I don't know if it really has been explored to get hard data.

How did you end up with a road with capacity for 2,000 cars and with 4,000 people to move, in the first place? What does "a road with capacity for 2,000 cars" even mean? The capacity of a road isn't measured that way. What does "with 4,000 people to move" mean? This is 4,000 people who have to travel this specific road at this specific time and have no other alternatives? And this is the only people who travel on this road? No one else travels on it?

I don't think your scenario is realistic, or even well-defined, so I can't answer it. And I can't even figure out what your proposed solution is. But to the extent your solution is to have a HOT lane, imagine if instead you replaced the HOT lane with a variable toll lane (same as the HOT lane, but without letting carpoolers travel for free). I can't think of a single scenario where that wouldn't be strictly better.

Maybe some small fraction of the population won't be able to afford the tolls. In most realistic situations, those people will just use different roads (or different lanes on the same road). But barring that, you could always give those people a fee waiver. Exempting that small fraction of those 4,000 travelers that are in poverty and carpooling and using the road to get to work will be much cheaper than exempting all travelers who carpool. I'd argue against exempting even them. If they are uniquely suited to a particular job that requires them to travel on a toll road every day, maybe their employer can cover their tolls to get there. Or a host of other possibilities. But to keep it simple, fine, what if you exempted them?

I could go on, but really I don't think you've presented a specific, reasonable scenario.

Road capacity is usually measured in cars per hour, and is about 2,000 per hour per lane. Of course it is not a capacity of any piece of road, though effectively it is -- at a 1.8 second headway, it's the number of vehicles that can pass any given point on the road. When you get congestion, the capacity drops -- there is a very bizarre sideways "U" shaped function used to describe the collapse. The capacity only varies a little with speed above around 25mph -- as long as flow is good. If flow goes bad it will drop -- and thus a traffic jam.

So, for the road in question, realize that 2,000 cars can go past a point on it in an hour, and that due to commute hours, there are 4,000 people who want to do that. This is true for all the points on the road (sort of) but of course people enter and exit at different points so it's not exactly the same.

So yes, those 4,000 people need to find a way past that impasse. There are options for them. Changing their commute hours. Taking trains. Taking alternate routes (which have their own capacity limits and loads) etc. But the best, in terms of use of resources is to get occupancy up to >2 people per vehicle. Then everybody goes when they want to, the road is not overloaded, nobody goes out of their way. So the government wants to make that happen. Or it wants to get people on the train, and it tries to do that of course but only goes so far.

Right now, the carpool lanes do not recruit nearly enough carpoolers in many places. So no, they aren't going to raise the prices on carpoolers. Though the real thing they should do is to find a way to only give the incentive to true carpoolers, who took a car off the road.

You say "use different roads" and sometimes people do that, but in many cases there are no choices that are anywhere near as direct as the road in question, and those roads have their own traffic too. In a sense, people "use different roads" over the very long term, by moving to a place where the commute is easier.

I don't believe a lot of people carpool for the money savings. As we've argued before, the incremental cost of their own car is lower than the full cost, but more importantly, people perceive it as being even lower. Many people imagine incremental miles as only costing "gas money."

The other driving factors for carpooling are carpool lanes, not paying for parking, company, and not having to drive -- which goes away if you have robocars.

How did the situation come about that only 2,000 cars can go past a point on it in an hour, but 4,000 people (virtually none of whom carpool) need to go past that point in an hour? What are these people doing currently? It doesn't make sense. They must be doing something currently. The only way I could see this happening would be if some other lane or route suddenly closed. If you have two lanes that can handle 2,000 cars each, and 4,000 people who use them every day to commute, and then suddenly one of those two lanes is unavailable, then yeah, that could happen. But that's usually a temporary situation, and I'm not sure inducing carpooling would fix it quick enough. (And even if it did, I could see using it in such an emergency situation, but not being a good solution for the general problem of congestion.)

A lot of people don't carpool at all. Carpooling doesn't currently make much sense. Maybe with robocars it will make more sense. Maybe even enough that carpool lanes will fill up. Probably so, since a robocar can easily pick up one person who wants a cheap ride and a second person who wants a fast ride, and then use the carpool lane. (Replace "one" and "second" as appropriate if the lane requires more than two people in the car.)

There are a lot of people who use public transportation for the money savings. The cost for them is not incremental, because they can't afford a car. (These, by the way, are the kinds of people for whom $1,000/year is going to be a big deal.)

It's just some numbers I threw out. Right now, because average car occupancy is 1.5 (though it is probably lower for the commute) if the highway needed to move 3,000 per hour it would do so within capacity. If the highway is regularly congesting, then it is trying to move more. Pick your number. The point is we have lots of regularly congested highways because more cars are trying to travel them than they can handle. (That's the source of the majority of congestion, though accidents and strange events also play their role sometimes, but they only cause major problems when the road closer to capacity.)

The point is, if you are a government managing roads, and you have congestion, you can solve it by building another lane, or by increasing vehicle occupancy. One of these is a lot cheaper. Though they usually do it by trying to do both at once.

And yes, they run into the problem of induced demand when they increase capacity, but my view is that internet based full metering should prevent induced demand from becoming a problem. If you cap the number of vehicles that can use the road, then the people will be forced to either get in under that cap (following whatever rules govern this) or finding alternate ways to to the trip.

If the maximum highway capacity is 2,000 cars per hour, and it is regularly congesting, then it is actually moving fewer than 2,000 cars per hour. Probably much fewer. So while there might be people in more than 2,000 cars per hour who want to use that highway, they don't need to, as they're already doing something else.

In any case, if the road is regularly congesting, I'm fine with implementing tolls and/or meters to keep it from regularly congesting. What doesn't make sense, in my opinion, is exempting carpoolers from those tolls. Giving something away for free causes shortages; it doesn't fix shortages.

If a government is managing roads, and they have congestion, neither building another lane nor increasing vehicle occupancy will solve it.

Metering might solve the problem. But if you exempt too many people from the metering, then it won't. Right now it's okay to exempt carpoolers from the metering, because carpooling is rarely feasible in most places. In the age of the robocar, it probably won't be.

The "something else" they are dong is having longer commutes. The problem of course is that a large fraction of people have to be at work at 9 and have limited choice.

It's overstating the induced demand concept to say that an extra lane doesn't solve it. It just doesn't solve it permanently, if you "allow" more people access to the road than it can handle again.

What governments currently do almost all the time is both add a lane and make it a carpool lane. In doing so, they increase the pre-carpool capacity of the road (in vehicles and people per hour) and they induce some carpools increasing the capacity in people per hour. Giving the carpools free access induces more carpools (one would hope) and that increases the capacity even more. It's a good deal, because it's a cheaper way to increase capacity than adding more road is. (They also factor in other externalities besides congestion, which is part of why electric cars get in the lane free too.)

As such, it solves the problem for longer. Not permanently, but more than just the new lane addresses it.

In addition, extra lanes are sometimes effectively impossible for physical reasons. (Except at the extreme cost of tunnels or overhead freeways.)

But these rules are about to change.

Yes, when I say that extra lanes (and carpooling, which effectively accomplishes the same thing though with different costs) don't solve the problem of congestion, I mean they don't solve it permanently. They are sometimes an effective band-aid. Sometimes not even that.

The best way to solve the problem is to stop subsidizing roads. That would be a permanent solution.

If the government gave away free food, the permanent solution to the long lines that would form would be to stop giving away free food.

(I'm not sure when extra lanes are effectively impossible. Not when you have the power of eminent domain. They might be expensive, but the expense is not charging a free market price for use of roads. Also, there are times when extra lanes are not effective, because the bottleneck is in a downstream location outside of the jurisdiction of the government that wants to build the lane. That's a situation where metering might be most effective, although this is still difficult because often there are federal and state roads that are alternatives to one another.)

P.S. If the "something else" is just having longer commutes, metering would completely solve the problem without any lane additions or carpooling. Everyone would have a shorter commute, though some might have to arrive earlier than they'd like.

The problem is the induced demand. Interestingly, that suggests a possible solution: Implement variable tolls, but give current regular users of the road a free pass (or better, a fixed, discounted price). Then only the induced demand has to pay the possibly high variable fees. This is also probably similar to how the free market would solve the problem. Demand-based pricing, but with subscriptions available, in limited quantities, for regular users.

It isn't so much the subsidy as it is the allowing unlimited access. For example if a private company owned a road in a mature area (where building additional roads is extremely difficult) they might decide to sell unlimited access to the road for an annual flat subscription fee. They might sell more subscriptions than there is capacity for. It would result in the same problem and might even generate maximum revenue because there is no financially viable alternative for the drivers or competitors.

But what is true is that it's far more likely for a government to allow unlimited access than a private owner, but that's beside the point for now. 99.9% of the road network is government owned and already exists. It will continue to be. I've read the analysis of the advantages of private roads, but it's theoretical for now, we must figure out plans for the world we have.

While there are certainly other ways to cause massive congestion on the roads, subsidizing a popular road is guaranteed to do so.

Government ownership of a road doesn't require them to subsidize it, though if the government stopped subsidizing roads you'd be much more likely to see private ones popping up.

I'm sure you don't have any direct control but I figured it might be helpful at least for you to know that I saw some pretty malicious ads after opening the Forbes page on a phone (mobile device). The ad(s) seemed to redirect to another page or at least hijack the entire screen and they included popups designed to look like 'official' popups, e.g. mentioning my phone device type.

I will track it at least and register some complaint, though I will need a specific example. Ads seem to go through cycles of doing absolutely anything advertisers want to being more respectful of readers.

Add new comment