Weapons autonomy is rocketing
By Heather M. Roff, Ph.D.
Best Defense guest columnist
While we debate whether or not it is a good idea, weapons are steadily becoming more autonomous, most notably in target identification.
That’s the core conclusion from a study I just completed in an effort to bring clarity to academic and policy debates about weapons autonomy. To do that, I set out to collect data and provide empirical evidence on the majority of weapons exporting countries’ arsenals. The logic was that the most well equipped would have some of the most sophisticated systems, and they would then try to recoup R & D and capital costs through exports.
For those methodologically inclined, my data (found for free here) covers 284 systems, most of which are all currently deployed, from the top five weapons exporting countries (the United States, Russia, China, France and Germany). While we do code for some emerging and developmental weapons, the majority of systems have been around for decades and have gone through various “upgrades.” Moreover, all data is drawn from public and non-classified sources. This includes specification sheets and websites of the weapons system manufacturers, reports from national defense organizations and NGOs, defense publications, such as Jane’s Defense, and websites aggregating publicly available information, like globalsecurity.org. That said, due to the secretive nature of some systems, as well as the difficulty of getting any information from some non-Western states, some systems had to be dropped from the data due to a lack of viable information.
We then coded each system for 18 different capabilities that it can “do” by itself. These range from homing and navigation, to target identification, prioritization, fire control, auto-communication, and learning and adaptation. It is also identified by manufacturer, date of deployment, domain (air/land/sea), and type of system (e.g., air-to-air-missile or unmanned aerial vehicle).
The results are telling. What we see is that “autonomy” in various functions is not a new thing. Indeed, if autonomy means the ability to undertake a particular task by itself (like self-driving), then it requires various capacities to enable it to carry out the task. Various autonomous abilities, like homing and navigation, emerged in the 1960s and 1970s. This, of course, makes sense when we think about the needs for countering nuclear threats. The United States required intercontinental ballistic missiles and missile defense, and so those systems needed the ability to find targets very far away from launch or to home in on and interdict incoming threats. As the technology matured and computing processing power and sensors got better, smaller, and cheaper, we see them deployed on more systems.
Moreover, beginning in the early 1980s and into the early 1990s, we see another major push into weapons capabilities. Here there are systems that can automatically sense and detect targets and fire upon them without a human needing to hit the “fire” button. These surface defense systems include land-based and sea-based varieties, but the key is that they are able to use various sensor technologies to identify, track and fire on a multitude of incoming threats without a human operator needing to be “in the loop.” This is not to say that they are all on this mode all the time. Indeed, these systems usually have the ability to be on direct control or semi-autonomous modes. The Patriot Missile system, for example, has three settings: direct control, semi-autonomous and “auto-fire.” In the semi-autonomous mode, it still requires a human to launch a missile, whereas in auto-fire it does this itself.
The most recent shift, however, of technological capability is not, as some might think, the face of a RQ-1 Predator. While many of the remotely piloted aircraft came into their own in the 1990s and early 2000s, and along with them the ability to loiter extensively, providing persistence in permissive air spaces, the emerging trend in autonomy has less to do with the hardware and more on the areas of communications and target identification. What we see is a push for better target identification capabilities, identification friend or foe (IFF), as well as learning. Systems need to be able to adapt, to learn, and to change or update plans while deployed. In short, the systems need to be tasked with more things and vaguer tasks.
So, while humans will undoubtedly be orchestrating combat from somewhere, the extent of what they know and how much they delegate to these emerging weapons systems is still to be seen. Depending upon the task, depending upon the required payload capacity, we could see a menagerie of systems compiled and reconfigured in a variety of ways.
Heather M. Roff, Ph.D., is a research scientist with the Global Security Initiative at Arizona State University, a senior research fellow at the University of Oxford and a fellow at New America.
Image Credit: Heather M. Roff
