True, it is possible to implement autonomous driving without change in the existing infrastructure, but his technology will be able to say more than that, one possible implementation of it is in forced lane management for controlling traffic, yes but it had more capabilities than that.
Using magnet is indeed a great idea. Challenges include installation of the magnets and the maintenance of them. The downsize is it doesn't provide more information than as a lane divider. Further information is provided by GPS, cellular signal and others similar technology. Since Apple iBeacon has been a topic on the day iPhone 5s was launched, will iBeacon come to rescue?
It seemed for a long time that autonomous car advocates had to pretend that V2I comms were not necessary, because including infrastructure changes into the equation would put the rapid deployment of "autonomous" cars in jeopardy. Looks like maybe carmakers are getting beyond the pretense.
Solutions like this one would make the whole autonomous car concept less reliant on bleeding-edge vision algorithms. And more to come, by the way. Credible techniques to accommodate other visual aids we depend on while driving, such as reading road signs, will need to be developed as well. Again here, Star Trek era vision algorithms would work in principle, to read road signs as a human does, or we can get real and create some kind of lower tech RF or optical comm solution.
It looks to be quite an ambitious project. It would be something like being in dreamland. The infrastructure needs to be just perfect and there would be no place of bad roads, bad driving conditions etc. Many big companies and entertainment parks can deploy this and would be of much use.