The Coming Future of Autonomous Drones

The inherent advantages of U.S. Medium Altitude Long Endurance (MALE) drones—like the Predator, Reaper, and Global Hawk—for spying on potential adversaries and attacking suspected militants has made them the default counterterrorism tools for the Obama administration. Unlike manned aircraft, or special operations raids, MALE drones can hover directly over hostile territory for between fourteen and forty hours (depending on their weapons and sensor payloads) without placing pilots or ground troops at risk of injury, capture, or death.

Presently, a constraining factor in how they are used is that they require more people to man unmanned aircraft than most people realize. For example, Predator or Reaper strike drones require between 168 and 190 people to maintain one combat air patrol (CAP), or the ability to sustain twenty-four-hour coverage of a specific area. This breaks down to roughly 30 percent with the mission control element, or pilots and sensor operators to control from a distant site via satellite communications, coordinators, and maintenance or administrative personnel; 30 percent in the launch recovery element, or pilots and sensor operators to control from the launch site via line-of-sight communications, and maintenance or administrative personnel; and 40 percent in processing, exploitation, and dissemination, or the video and signals intelligence analysts, and maintenance or administrative personnel. For the Global Hawk spy drone, as many as three hundred personnel are required for a CAP.

The military’s strategy to make the Predator, Reaper, and Global Hawk systems more efficient and less costly is to replace human beings with greater autonomy in how drones operate. The idea of autonomous drones is understandably alarming to many, especially in making life or death decisions to drop bombs. This fear, however, should be tempered somewhat by the fact that fewer than 5 percent of military unmanned aircraft are capable of dropping bombs. Moreover, through dozens of conversations with Pentagon and Air Force staff and officials who control, command, or develop future concepts for drones, I’ve learned that there are no plans for fully autonomous drones that conduct lethal strikes (although some believe that in threatening air defense environments, MALE drones should be empowered to autonomously return fire when attacked by air-defenses missiles or jet fighters).

Last week, the Defense Science Board—an advisory panel that advises the Pentagon on science, technology, and special issues—published a July report, The Role of Autonomy in DOD Systems, which addresses some of the new strategies and concepts for greater degrees of autonomy within the U.S. military. There are two findings in the report worth highlighting. First, the report presents three ideas for lowering the personnel required per CAP. To note, there is no mention of automating decisions to use lethal force.

Using the Air Force staffing requirement for its remotely piloted aircraft combat air patrols (CAP) as an example, possible ways autonomy can reduce manpower include:

— Piloting/sensor operation: Currently, it takes multiple operators to manage the flight and sensor operation functions for both the Predator and Global Hawk systems. While it will be essential to maintain a “human in the loop” to supervise operations and to make critical decisions such as those related to weapons release, the effective use of autonomy technology will enable a single operator to manage multiple vehicles. This goal can be achieved by delegating decisions such as take-off and landing, waypoint navigation and sensor-enabled situational awareness to the computer.

— Maintenance: Using autonomy for on-board equipment health and status monitoring should improve reliability and reduce the maintenance staff required to support operations.

— Exploitation: About a third of the staff required to support Air Force UAVs are devoted to processing sensor data and exploiting them to create useful information. Even with this staffing level, the rapid growth in data volume is making it very difficult to keep up. There are many opportunities to use autonomy capability to increase the capacity of the intelligence analysts assigned to the exploitation function.

The second section assesses how autonomous drones could be used by potential adversaries of the United States. Since the United States (and to a lesser extent Israel) dominates the global unmanned aerial system research and development for both surveillance and strike drones, this could emerge to become a minor threat to the homeland or deployed U.S. forces. However, many states are now rushing into the field. For example, estimates of states with active drone programs range from forty-four to seventy, accounting for roughly 680 drone programs around the world—a significant jump from 195 in 2005. The report lists four scenarios for how autonomous unmanned systems could be used against the United States.

Direct Attack on CONUS. UxVs (unmanned “systems” comprising: UAVs, UGVs, UUVs and USVs), could be used against CONUS for the same reasons the United States might choose them in the reciprocal case—reduced friendly casualties; increased availability of systems in the battle area (lack of human physical limitations allows air-refuelable UAVs to stay airborne longer or fight from a greater stand-off range, thus enabling more average aircraft on station for a particular task over time at equal cost); increased instantaneous force size (assuming a cost advantage to the equivalent manned system), etc. Enablers for their use would be characteristics like stealth or threat jamming systems—the same as exist for U.S. systems.

Defensive challenges to the U.S. would be to overcome whatever numerical advantage, if any, the UxVs afforded the adversary. U.S. forces would have to be sized to meet the threat. A conflict with high attrition rates would eventually favor the force with more UxVs if manned system training requirements exceeded the time needed to manufacture unmanned systems. Basing would be the key limitation for a peer enemy. Unless the enemy successfully fields both carriers and a carrier-based UAV, he will likely be limited to either large/refuelable long-range systems, or to operating from some small number of known and observable local bases (e.g., Cuba). This factor alone is likely to limit attack effectiveness against CONUS…

Overall, we do not anticipate enemy UxV forces significantly changing the calculus for this scenario in the near or mid-term.

U.S. Attack on Adversary Homeland. The adversary has many UxV options to oppose a U.S. attack. As in the parallel CONUS case, basing is the largest issue for the U.S. It is generally accepted that land bases within moderate range of a peer adversary will likely be closed, or at least badly degraded, in the event of a conflict. In the Pacific, at least, this has motivated new initiatives in long-range carrier-based attack, relying on carrier mobility and location uncertainty to sustain survivability. Consequently, an adversary might well structure UxV forces to support targeting of our Carrier Strike Groups. Given the range of aerial, surface and subsurface candidates—as well as dispensable sensors—available this would be a very serious threat, and it could also plausibly include non-trivial organic attack capability. This threat could be extended to rear echelon supply convoys and other combat support assets which have not had to deal with an airborne threat in generations. Higher levels of on-board autonomy would circumvent our abilities to degrade UxV performance by simply jamming the adversary C2 links. With respect to actual defense of the adversary’s homeland, the adversary’s choice of UxVs would more likely be based on its ability to generate (and sustain) a larger instantaneous force size at lower cost (assuming adequate tactical performance) to make the U.S. attack as expensive and difficult as possible...

Integrated defense system would substantially complicate U.S. mission planning and execution. Unmanned platforms lost to attrition might also be easier to replace than manned platforms. Close in basing would remove any enemy dependency on satellite communications, thus removing a significant degree of vulnerability.

Regional Warfare (i.e., attacks on U.S. forces outside the CONUS). The scenario considers one similar to that faced by the United States during the Korean War, in which U.S. and a symmetric adversary or an adversary client’s forces were fighting on third-party territory, and the adversary homeland was a sanctuary. The adversary would likely see the value of UxVs in a similar manner to its view of UxVs’ utility in attacking the CONUS—fewer casualties, larger instantaneous and average-available forces in the theater and opportunities to find and engage U.S. naval forces. In this case, however, the adversary would expect to have sufficient basing, as well as home sanctuaries from which to sustain its UxV fleets. The training cost argument could be less compelling if, as might be expected, the number of forces deployed in such a conflict was small compared to the total available in garrison. If the distance from adversary homeland were short, UxVs could be operated from sanctuary directly into the battle space. This would enable application of large numbers of small systems, conferring significant advantage if they were capable of surveillance, attack and/or electronic/cyber warfare. Defending against large systems would pose the same challenges as in the case of an attack on CONUS. Defending against a proliferation of small systems while operating our own manned and unmanned systems would be complex and represent a serious threat.

Actions Short of Active Warfare to Gain Military Advantage in Case of Hostilities. In this scenario the value of UxVs to the adversary would be to extend, supplement or replace the capabilities of overhead systems to provide extended range detection and tracking of U.S. forces. UxVs for these purposes would not have to be stealthy and would not need strike or defensive capabilities so the cost would be much less than manned systems for equal coverage. Broad Area Maritime Surveillance-like systems for tracking U.S. ship movements would be an effective way for a near-peer to force carrier battle groups to stand off significant distances. They could also track troop buildups and harass supply lines. The combination of space-based queuing assets with long-range, long-endurance surveillance UxVs could add substantially to U.S. attack vulnerability at the outset of conflict.

In his April 2012 speech that marked the first official U.S. government acknowledgment that American drones were conducting targeted killings, senior White House counterterrorism adviser John Brennan conceded, “We are establishing precedents that other nations may follow.” Focusing primarily on technology and systems integration issues, the Defense Science Board report did not mention that U.S. decisions to reduce human interaction with drones could have a normative influence on other countries that pursue drones. Nevertheless, before the United States goes down the path of further automating the use of drones in battlefield and nonbattlefield settings, there should be public discussions and congressional hearings—a debate that has never occurred despite the escalation of targeted killings in America’s Third War.