I'd imagine sensor fusion in this context isn't a singular problem right? It seems entirely plausible that for Tesla's system, it's straightforward to integrate cameras where the resolution is known and data is homogenous in a certain way, and that it's hard to integrate and use radar data into that system.