Reversing the U.S. Navy’s trend toward becoming a brittle fleet of a small number of expensive high-end platforms will require innovation and ingenuity.1 Human-machine teaming can move the Navy in that direction by allowing limited artificial intelligence (AI)-controlled unmanned systems to act as “loyal wingmen” to manned platforms and eventually to take over more of the dull, dangerous, and dirty work. But that will require innovative command, control, and communications (C3) architectures to enable computers (the fourth C in C4) to receive proper guidance from human commanders, who must learn to trust their autonomous machines. Determined leadership is required to overcome the technical challenges and bureaucratic resistance that stand in the way of transforming a brittle fleet in to a formidable 21st-century fighting force.
Toward A Brittle Fleet
Big ships are more efficient than their smaller counterparts in carrying ordnance, power, fuel, and equipment. There is a natural tendency to take advantage of space and pack more capabilities into each ship. With so many capabilities invested into a platform, commanders and builders then seem compelled to add significant defenses, further increasing the cost. A vicious cycle is perpetuated: as costs rise, the number of platforms that can be acquired declines; as the numbers decline, each platform becomes dearer, calling for more costly defenses.
This tendency to go big is reinforced by the desire to reduce the possibility of losing sailors. Focusing on the survival of individual ships, however, can compromise the fleet and the mission. Concentrating the fleet’s capabilities (particularly offensive ones) into a limited number of platforms allows enemy commanders to track a smaller number of ships and design an attack to which we can provide only a limited response. It is impossible to build an invulnerable ship. If one of these ships is attacked successfully, its multimission capabilities are lost to its commander.
Numbers matter. If the fleet is too small, commanders faced with a near-peer competitor will be reticent to deploy within range of the competitor’s capability, thus forfeiting the ability to maintain presence and influence events in the region. In deterrence operations, fleets with larger numbers increase a competitor’s requirement to track opposing forces.2 In combat, the number of weapons an enemy commander must expend to have the same confidence of annihilating a larger opposing force is significantly greater than against a smaller force.
A fleet with large numbers also can better enable wars of deception, maneuver, and local initiative. In the past, one solution was to employ a mixture of high-end platforms and large numbers of low-end platforms. In World War II, the high end was represented by the Essex (CV-9)-class fleet aircraft carriers and the Baltimore (CA-68)-class heavy cruisers. The low end was represented by the Casablanca (CVE-55)-class escort carriers and the John C. Butler (DE-339)–class destroyer escorts. These smaller destroyers acquitted themselves quite well in the hands of Commanders Ernest Evans and Robert W. Copeland.
Unleash Unmanned Systems
Autonomous platforms partnered with manned systems can provide a solution to the need for numbers and overcome the hesitancy arising from the potential risk to individual manned platforms. The U.S. Navy recently created the position of undersecretary for unmanned systems to develop a strategy for the use of unmanned vehicles, including autonomous platforms, in multiple domains. Today, the U.S. Navy has developed and is experimenting with several unmanned platforms that could potentially provide the numbers to make the fleet less brittle (more robust) and perform many dull, dangerous, dirty jobs—relieving sailors to focus on fighting as team leaders in the formation.3
The antisubmarine warfare continuous trail unmanned vessel (ACTUV) project, funded by the Defense Advanced Research Projects Agency, launched in early 2010 to develop an antisubmarine drone (unmanned surface vehicle). Autonomous ships such as ACTUV could be employed to support manned ships as armed escorts with antiship cruise missiles or multipurpose missiles. These unmanned drones could be reconfigured to conduct mine clearance or mine delivery operations.4 They also could be used as long-endurance decoys to draw enemy fire away from manned and critical platforms. Systems such as the Pentagon Office of Force Transformation’s M-80 Stiletto or the Juliet Marine Ghost boat are examples of small, optionally manned platforms that enable the concept of distributed lethality. Operating optionally manned platforms in swarms with multimission ships would reduce the risk to manned platforms.
The MQ-25 Stingray unmanned carrier aviation air system program—formerly the carrier-based aerial-refueling system—is focused on providing carrier-based refueling capabilities for the carrier air wing. It is being considered for limited surveillance and reconnaissance missions. In the future, such autonomous aircraft easily can be developed to act as loyal wingmen to manned aircraft, carrying weapons and sensors to better enable the force to fight.5 Such unmanned platforms can take on risk to close the enemy, identify targets for the manned platforms to engage, and act as airborne magazines for the force.
In addition, the Navy is experimenting with unmanned and autonomous undersea vehicles (UUVs/AUVs) for a range of missions. They are being considered to support submarines or operate on their own to conduct antisurface and antisubmarine warfare. However, such platforms also could be used to support more significant missions such as nuclear deterrence, bolstering the maritime leg of our nuclear triad. Just as with conventional deterrence, the number of platforms employed for nuclear deterrence matters. The Navy will decrease the number of ballistic-missile submarines from 18 to 12 when it deploys the Columbia class to replace the aging Ohio (SSBN-726)-class ballistic missile submarines (this includes the four Ohio-class submarines converted to carry cruise missiles instead of nuclear ballistic missiles). This decrease will have a significant impact on the deterrence effectiveness of the force, because larger numbers of potential targets greatly reduce the probability of the enemy locating and maintaining track on enough SSBNs to negate their deterrent effect. To regain the effectiveness of this critical deterrent force, the Navy should employ high-endurance AUVs not to carry nuclear weapons, but to emulate and act as decoys for the SSBNs and create doubt in the minds of potential adversaries. Adding numbers would force adversaries to deploy more assets to effectively locate and track this leg of our nuclear triad.
There are still many missions that cannot be turned over to AI. Sailors and Marines, for example, will continue to conduct boarding operations of suspect ships. We can, however, ensure that armed AI-enabled ships and aircraft provide over watch support for boarding operations. Such divisions of labor must be explored and fully understood to be effective. Just as important, imagination must be unleashed and enabled to explore other unmanned opportunities.
Build Loyalty and Trust
Warfighting machines must be truly autonomous yet cooperative—behaving in a manner that engenders trust in their human commanders by remaining loyal. Today, remotely piloted vehicles (RPVs) such as Global Hawk and Reaper have delivered great capabilities—particularly endurance—but have not provided the promised cost savings. While the pilots have been removed from the cockpits, the increased endurance of these RPVs require more pilots per platform. The continuous data links needed to control the RPVs are costly, and the dependence on satellites represents a critical vulnerability.
New C4 architectures, such as Matrix Networks, remove single vulnerabilities but still require strong encryptions and careful coding to ensure commanders remain in command and the systems remain loyal.6 At the same time, the AI engines used in these systems must be designed to operate in communications-degraded or -denied environments. Action logic matrices must be developed to switch the machines to an engagement mode where they respond to short messages coming across secure low-bandwidth channels. We employ such techniques today in the form of doctrine “play books” to inform human operators on ships and aircraft.
While our competitors have developed, and continue to develop, communications- and sensor-jamming systems, their efforts will be more effective in interfering with remote-controlled systems than in disrupting the terse commands that activate doctrines in advanced unmanned systems. In the case of an AUV, communications requirements must be greatly decreased because they operate underwater.
Operating autonomously will require continued development of expert AI systems. While not yet approaching the AI capabilities of 2001’s HAL-9000, sophisticated AI systems are in use today. For example, commercial stock trading uses expert systems for trading at extremely high speeds. They also are used in additive manufacturing for extremely small and detailed construction. The Navy uses a form of expert AI in the Aegis combat system—especially in the missile’s terminal modes, where communications are impossible because of range and closing speeds—and in modern antiship cruise missiles. Humans are not in the loop in some of these modes. Computers can do impressive tasks that exceed a human’s ability to keep up or pay attention. On the other hand, they also can cause unintended consequences, such as the recent flash devaluation in the London stock market, the result of a computer-generated sell-off. More recent advances in the scientific AI realm foreshadow advances arriving far sooner than many expected.
To take advantage of the coming technical sea change, commanders and developers must start to war-game doctrines under which they want these systems to operate. Scenarios and questions must be developed in advance. How do we want an armed unmanned platform to respond when communications are lost? How do we respond to signs of hacking? How should systems react to defend themselves in peacetime? How do they act when someone is attempting to board them? How should they respond when in contact or conflict with unmanned systems employed by competitors and adversaries? These are just a few basic questions; more complicated questions must be addressed in the context of modern conflict. A series of war games should be conducted to better understand the problems and to develop a preliminary doctrine to pave the way for an unmanned revolution in the fleet.
Expand Unmanned Autonomy
Moving toward unmanned enhanced fleet structures requires the unmanned community to realize its promise of providing systems that save the force money, mostly from reduced personnel costs. Developers must minimize the number of personnel tied to unmanned systems and prove that AI systems will operate as intended in complicated environments and perform complex tasks. Commanders must delegate actions to these autonomous systems if their full value is to be realized, but they will grant such trust only if they believe these systems are truly effective.
Commanders, experts, and operators also must be more imaginative and innovative in how to employ unmanned forces and the range of missions they will be allowed to perform. The Navy must explore the risks and rewards of using these systems and develop the calculus to understand them. To have AI-enabled systems behave properly, the Navy must know the range of tasks and scenarios they will be expected to perform and develop the logic table for them to follow. We must be careful to not underestimate what unmanned war fighters can bring to the fight.
Given the lethal precision strike that modern warfare demands, extensive employment of autonomous unmanned systems is key for the Navy to avoid becoming a brittle fleet. The Navy must explore a wide range of roles and missions for the unmanned force or it will be forced to play catchup with the rest of the world.
2. John Schank, et al., “Sea-Based Strategic Defense: Analysis of Alternatives,” RAND Report MG-932 (February 2010), 110–120.
3. Tomas Kellner, “Someone’s Gotta Do It: This Collaborative Robot Does the Dull Jobs Few Humans Want,” GE Report, 21 January 2015.
4. Remarks of Mr. Charles Werchado of OPNAV N81 September 2016.
5. Statement of Deputy Secretary of Defense Bob Work, 30 March 2016.
6. Jeff Cares, The Foundations of Network Centric Warfare (Alidade Press: Newport, RI, 2005). See also Alexander Bordetsky, Stephen Benson, and Wayne Hughes, “Mesh Networks in Littoral Operations,” USNI Blog, 12 May 2016.
Commander Pournelle is the senior director for gaming and analysis for the Long Term Strategy Group, LLC. He is a former Navy surface warfare officer and operations analyst. He recently retired from the staff of the Department of Defense Office of Net Assessment.
Photo Credit: General Atomics