It comes down to space. Cars occupy way more space per passenger than trains do. A fully loaded subway train can easily carry almost thousand people. The 7000 series in the DC Metro carries ~175 people per car maximum, and the trains are 8 cars long, for 1400 people per train. Normally, 26 trains per hour are scheduled during rush hour, allowing up to 36,000 people per hour if all trains were 8 cars long (they currently aren't due to power constraints). Other systems that have longer trains, better headways, and better seat configurations have way higher capacities. New York for instance could achieve much greater capacity given its longer trains and a modern signalling system.
By contrast, a road lane (which uses the same space) currently typically carries around 1000-2000 cars per hour. In ideal circumstances it can be almost 5000 cars per hour. Even assuming quadruple capacity from self driving cars (not realistic), 20,000 cars per hour in ideal circumstances would compete with trains at a far higher environmental cost.
But in more realistic scenarios with quadruple capacity, 4000-8000 cars per hour would fall way short of a train. Given that we have to account for pedestrians, cyclists, etc., I can easily see traffic lights continuing to exist, making self driving cars not achieve the miracles people expect.
By contrast, a road lane (which uses the same space) currently typically carries around 1000-2000 cars per hour. In ideal circumstances it can be almost 5000 cars per hour. Even assuming quadruple capacity from self driving cars (not realistic), 20,000 cars per hour in ideal circumstances would compete with trains at a far higher environmental cost.
But in more realistic scenarios with quadruple capacity, 4000-8000 cars per hour would fall way short of a train. Given that we have to account for pedestrians, cyclists, etc., I can easily see traffic lights continuing to exist, making self driving cars not achieve the miracles people expect.