With so much attention being devoted to commercial drones these days, it's easy to forget that the military's long-term strategy for unmanned systems is still a work in progress. A lot of the funding for drone operations comes from the special budget for the war in Afghanistan — and as the drawdown begins, that money is going to dry up.

That's a big problem for the military. It now has to transform what was originally a stopgap solution into a sustainable function of the armed services. That process is also going to grant drones a more important place in combat, according to the latest revision to a 25-year roadmap the Pentagon released this week. Depending on the kind of drone we're talking about — unmanned aerial systems (UAS), unmanned ground systems (UGS) or unmanned maritime systems (UMS) — the Pentagon foresees an array of different problems that all need to be tackled over the next quarter century.

The near-term will mostly involve making sure the technology works at all. As we move into the next decade, that will give way to exploiting the real potential of unmanned systems. That's when a lot of the changes to doctrine and strategy will begin to take effect.

Advances in commercial technology will help speed that process. So will research being done by the Pentagon itself. But the biggest challenges aren't necessarily technical.

Take how humans will be expected to interact with their machines in the field. While it's tempting to think about whole fleets of unmanned ships steaming across the ocean as one, the reality is that people are going to be right alongside them for the foreseeable future. There may be fewer humans than drones in a unit. But you might have, for instance, a human pilot being aided by a set of unmanned wingmen in the air. The military already has a name for this: Manned-Unmanned System Teaming, or MUM-T. Figuring out how much interaction will be needed is going to be an ongoing process.

Another problem strategists are predicting is the information overload that comes along with vast amounts of data. Drones right now are great at feeding live images and video back to their operators, but that creates obstacles of their own: The more unmanned systems the Pentagon operates, the more bandwidth they require, and the more money the Defense Department has to spend on commercial satellite contracts to carry those transmissions.

What's more, sending pictures back for people to look at is an inherently inefficient way of doing business. It would be much better, the Pentagon reasons, if the drone were smart enough to ping somebody only when it recognized an object of interest. That would cut down on the amount of data flowing back and forth, making it easier to load the system with more drones and freeing up people to do other things.

Grappling with that will mean having to sort out what we mean by "automation" — when human pilots are no longer directly controlling the drones remotely, as they do today. There are two ways to think about that: execution and performance.

"The difference between execution and performance," the report notes, "is that the former simply executes a preprogrammed plan whereas performance is associated with mission outcomes that can vary even during a mission and require deviation from preprogrammed tasks."

It takes a smart machine to recognize and respond to changing conditions.

Part of the solution lies in building drones — and equipping them — in adaptable ways. Weapons of the future won't be repurposed from helicopters but designed specifically for drone fleets themselves. These drone-specific munitions might be capable of different yields that can be adjusted before use; they might be ultra lightweight; or they be able to adjust their own trajectory in mid-flight.

Even further down the road — into the 2020s and 2030s — the military expects to learn more about nanoparticles and their potential for creating much larger explosions than today's ordnance. The Air Force is already experimenting with an aluminum powder that may prove promising, according to the report.