I think they are way off the mark as to the nature of the problem they are trying to solve as well as the timing of it. I believe there is a good possibility that we will have not thousands or tens of thousands, but millions of smaller drones in the air by 2020. And, yes, they will be autonomous or semi-autonomous. Putting Google cars on the road is harder than making drones autonomous.
The key to understanding here is that these are just robots. As we move further into the age of personal robotics, there will be many many tasks that a robot that can fly will be better able to do for us than a robot that is limited to walking or rolling around. Many of these devices are also very small. Once they become quieter and smart enough to auto-fly through hallways and crowds, I see no reason why these devices wouldn't go everywhere we go. I foresee them flying through doors into buildings, possibly switching into and out of rolling modes and delivering items right to a person, not just to a building. Or, to get away from the delivery theme, they could be flying around picking up trash, washing windows, getting leaves off of roofs, trimming trees, stringing poles so that nobody has to climb up them, changing lights on towers, flying the rounds of a security guard (even through halls),,, who knows, disposable cameras might even have the ability to sprout a prop, fly off 15 feet, take your picture, come back and land on your hand (there was a bracelet that did something like this on YouTube recently). It's all moving fast enough now that none of this is unrealistic.
So, given that it would take the government 15-20 years to deploy a system, it will be way too late. The need, the explosion of devices, will come from the home-based, personal uses, not commercial businesses, and it will come in the next few years. I've already seen devices that automatically launch, perform a chore, dock, and recharge. It's one of the next big things, and it is way closer than people think. If companies don't do it for us, we're going to do it ourselves.
These systems are going to have to control their traffic the same way people do, with eyes, ears, and some rules of the road. They are going to be interacting and intermingling directly with us, not just each other. Even thinking about centralized control is just a way to subsidize some scientist who would do better economically spending his time designing these devices himself instead of telling others how to control them. We're not waiting for his advice.