Sophisticated sensing systems are too heavy for the flybots, never mind the computer equipment to actually use the resulting data to plan routes. Instead they take a page from insect strategies (thereby combining the fields of unstoppable machines and multitudinous vermin in ways that could certainly never go wrong.) Instead of a single smart bot the researchers plan to unleash swarms of mass-produced machines which can completely cover any indoor environment through sheer numbers.
The double-rotored robots literally bounce their way through the building, warning each other of obstacles after the fact while passing information through the spreading network back to the source. In this way the most complicated construction can be searched quickly as long as you add enough bots. Lightweight mics and cameras provide feedback from all over the building, and once the information's back at the base you can use hardware as heavy as you like to reconstruct a complete map of the entire area.
Current applications are search and rescue, but the "boatload of bots" approach could work extremely well in off-word efforts. Instead of one or two extraordinarily expensive probes, where the only response to one of a thousand possible failures is to punch the air, mutter "Drat!", and try again ten years later, mass-produced probes of a few differently-equipped flavors could blanket an entire area. This is perfect for present exploratory missions whose whole point is usually "Can we find X Y or Z up there?" And if a few dozen probes break or blow up, who cares? Plenty left.
