I wish people would consider basic facts of geometry before posting inane ideas like "ripping up tracks for autonomous vehicle lanes." Especially on this forum, where we're supposed to be a bit more thoughtful than the typical Herald, etc, contributor.
A rough rule of thumb is that you cannot pump more than 2,000 cars per hour per lane down a highly optimized right-of-way. This rule is derived from the fact that you need some separation between vehicles in order to appropriately recover from emergency situations. The rule of thumb taught to drivers is "two seconds worth" of spacing. Following that rule religiously would result in a maximum of 1,800 cars per hour per lane NO MATTER what the speed limit is (at highway-type speeds). And that's optimistic, assuming no breakdown in travel flow due to congestion. In practice, most people cheat the "two second rule" but also most traffic flows breakdown somewhere above 2,000 cars per hour per lane.
Autonomous cars do not allow you to break the "n-second rule", just reduce it. Computers have reaction times too -- plus it is necessary to stay within the parameters of comfortable and feasible acceleration and braking power for a motor vehicle. And I have no reason to believe that traffic flows won't break down based on congestion over a certain threshold even if computer-driven; it just might be a higher threshold.
In addition, people do not want to just travel on these ways, they want to actually go places. That means getting in and out of the vehicle at some point, which takes time. Or having the vehicle enter and exit the traffic flow at some point. These are disturbances that make it increasingly difficult to go above the 2,000 cars per hour per lane flow. It might be possible to achieve 2,500 or even 3,000 cars per hour per lane in some special cases, such as long tunnels. But that's about it.
The reason I bring all of this up is that 2,000 cars per hour per lane is REALLY REALLY LOW considering the potential capacity of the right of way.
Even if you assume that everyone is car-pooling with 5 people per vehicle (fat chance) that only adds up to about 10,000 people per hour per lane. More likely, you would see no more than 3,000 or 4,000 people per hour per lane.
Now compare that with what even a primitive train system is capable of handling: you could easily have a train with 1,000 people arriving 15 times per hour, with plenty of time to unload and reload passengers. That's 15,000 people per hour per track without even stressing anything -- using the same technology that the Red Line uses. Using the latest in signalling technology, RER-style systems can transport upwards of 50,000 people per hour per track.
That kind of performance simply blows any kind of individual vehicle system out of this universe. Based nothing more than on the difference between having multiple independent vehicles vs multiple vehicles traveling as a unit. It doesn't matter if there is autonomous control or not. Remember: driverless trains have existed since the 1960s and the safety concerns apply just as much to them as they do to cars.
Another reasonable alternative might be autonomous buses that travel along a trunk line to combine for high capacity. I think that this will be a useful tool in the box of future cities. But, it's still not so easy to convert a train right-of-way into a bus trunk right-of-way at high capacities. Again, boarding and alighting is the bottleneck. Most high-capacity BRT are designed with multiple bus lanes and pull-out areas because it takes longer to get passengers on and off buses than it does for trains. Trains are naturally suited to having large numbers of doors open simultaneously onto level boarding platforms -- a very important requirement for achieving high throughput. Now that's not to say buses cannot achieve this -- they can, and they're getting better every year. Maybe some day the difference will become negligible.
At some point, the technological differences between buses and trains may converge: buses that draw electric power from overhead lines (or similar), that operate on guideways using computer or mechanical assistance, that are extended to 120' or more, and that have multiple, wide doors with smooth level boarding. At this point, you're almost starting to have a bus system that looks like Montreal's Metro. This is all great, and could be part of future systems.
But I'm talking about carrying capacities that are on the order of 20,000 - 50,000 or more people per hour per "lane/track". The idea of trying to replace that kind of capacity with something that can only handle 3,000 or 4,000 people per hour per lane is laughable, if not downright foolish. In fact, it's so foolish that it sounds an awful lot like an ideologically driven agenda. Much like Elon Musk's lame attempt to throw up "whoa look! Hyperloop!" or the idea of autonomous electric cars as some kind of replacement for HSR, I'm suspicious of people who think that magical technology can somehow contradict basic facts about geometry and reality.
