There are a couple differences between other centrally controlled systems and the one proposed by MIT. The main thing is that MIT’s system is expected to operate at extremely tight tolerances and takes the place of decision making at the vehicle level. Because of the high speeds these intersections will operate at, failures will be a big problems and there will be no human to step in to stop collisions especially if the central system is hacked to purposely cause collisions. In current centralized systems being used there is at least human common sense and caution operating to prevent carnage. This leads to the broader point that intersections such as modern roundabouts don’t require any controls and would only be optimized by decentralized communications, but would not require any communications to navigate. VANET style mesh networking between vehicles to coordinate speeds and lane changes would allow the vehicles to close gaps between each other and increase efficiency. Roundabouts are already more efficient than perpendicular intersections and would only be enhanced by autonomous vehicles becoming more common.
The second issue is that anyone proposing a new system should not be banking on an outdated paradigm like a centralized control scheme. MIT must because the intersections they are proposing are so complex and they like to show off real world tests of their system at work. Decentralized networks aren’t yet able to handle the complexity they need to hit the efficiency they are after. The future is decentralized and even MIT proposes decentralized communication for low volume intersections. We should not be building new infrastructure that needs complex communications to operate. I have a publication in the works that deals with decentralization, but for now I will say I am suspect of anyone in this space who thinks central control has any place in our future.