Software: The Key To Swarm

Paul Scharre has a great piece over at War On The Rocks on the future of swarm technology, and specifically counter-swarm. The piece masterfully illustrates how certain military “jobs-to-be-done” can, even today, be accomplished by drone swarms — from platform attack to ISR. Sharre focuses this piece on counter-swarm, and as he gets in to the weeds, things get interesting from a software perspective:

Swarming will require new concepts of operation and several paradigm shifts for the U.S. military. We must move from a paradigm of driving and piloting vehicles to mission-level command of a swarm. We must shift from waiting for “full autonomy” to seizing the opportunities afforded by operationally-relevant autonomy today. Instead of focusing on capability, we must focus on capability per dollar, and the advantages that mass may bring. We must shift our focus from survivability of individual platforms to resiliency of the swarm as a whole. We must think of not only payloads over platforms, but also software over payloads. Finally, instead of thinking about “unmanned” systems replacing people, we should think about robotic swarms as yet another tool to help warfighters adapt to a changing world to do their jobs — to fight and win.

What is imporant here is his focus on “software over payloads”. Why? Because in future drone engagements (undersea, air, or ground), Boyd’s traditional E-M theory of maneuverability as it applies to single platforms won’t actually be the discriminator for victory, except perhaps at a macro level. Drone combat won’t be a set of one hundred drones facing off against one hundred counterparts, by pair. Instead, it hundreds of assets facing down a similar assemblage. Each drone may be engaging multiple enemy drones at any given time in a scrum that looks (in volume — not in syncronicity) like competing flocks of starlings inhabiting the same airspace. Thus, the critical maneuverability will be that of the swarm in the aggregate, and its ability to re-task, microsecond-by-microsecond, its assets to effectively counter the enemy.

Building the capability to effectively wield a swarm is likely to be a processing-power-intensive challenge, as the software architecture will need to, in near-real-time, evaluate the simultaneous location of every friendly and enemy asset, calculate their speed, angle of attack, payload, potential target, etc., and then generate a momentary “strategy” for the entire swarm, moment-by-moment. Most importantly: drone swarms that rely on human operators are likely to be overmatched by their machine-driven counterparts (which is why Scharre emphasizes “mission-level command”).

And how to control it all? Software. Software is eating the war.