The great diversity of challenges that threaten the U.S. Navy in the near term—spanning all domains and levels of warfare—has led to an unequal concentration of effort, leaving some critical areas largely unaddressed. Maritime logistics ranks chief among these. The Navy should focus some of its investment, development, and integration in artificial intelligence (AI) to enable strategic and operational logistics at the high end of maritime conflict.
Promise and Peril
As is true of so many revolutionary technologies, AI offers both potential and risk across a broad swath of applications, suggesting better, faster, and more autonomous decisions in support of networks, command and control (C2), unmanned platforms, sensors, and weapons. But to fulfill this promise and manage the risk, each specific application of AI must be tightly focused on recognized challenges and rigorously tested for efficacy.
The resurgent threat to joint maritime logistics is as consequential as it is underappreciated, richly deserving the focused application of emerging technology such as AI. Great maritime powers have always sought the capacity to interdict their enemies’ vital lines of communication while protecting their own. Fleet Admiral Ernest King’s strategic retrospective still applies: “It is no easy matter in a global war to have the right materials in the right place at the right times in the right quantities.”1
The most significant threats to convoys at sea are submarine-launched torpedoes and antiship cruise missiles (ASCMs). Though torpedoes are the most reliable ship killers, they must be fired from relatively close range, while ASCMs trade some lethality for much greater standoff. Chinese conventional submarine loadouts for the anti-surface-warfare mission “are light on torpedoes and heavy on ASCMs.”2 To survive, a convoy must be able to detect and defend against both threats and then engage submarines, even out to the limits of ASCM range—from 23 nautical miles (nm) for the ubiquitous YJ-82, to as much as 270 nm for the more potent YJ-18.3
Compounding this daunting tactical threat is a lack of escorts. The head of the Department of Transportation’s Maritime Administration, retired Navy Rear Admiral Mark Buzby, reported to Congress that the Navy told him: “You’re on your own; go fast, stay quiet,” and hope to stay undetected.4 This is a vain hope, however. Modern all-source intelligence has made convoys far easier to find, fix, and track; space-based imagery can help refine these tracks.
The outlook may appear bleak, but an artful combination of force structure, tactics, and technology can protect vital sea lines of communication against even the most determined adversary. Five interrelated areas, each one enhanced by AI, support this new operational concept:
• Robust and covert networks
• AI-assisted command and control
• Unmanned platforms
• Networked, distributed sensors
• Networked, distributed weapons
Communication Is Fundamental
The dispersal of units suggested by distributed maritime operations complicates reliable communication while also raising the stakes should the network fail: Without communications, the fleet becomes divided rather than distributed. Absent mutual support, it risks destruction in detail. But there is an optimal middle ground between transmitting at full power from every emitter on board and huddling in the electronic darkness under total emission control, bracing for impact.
AI-aided decisions supporting an adaptive use of the electromagnetic (EM) spectrum can help optimize this balance. Like an “electronic idiot-savant,” existing AI excels at solving complex but well-defined problems such as dynamic communication link maintenance.5 A convoy’s network must be covert enough to be generally undetectable, reliable enough for low-bandwidth C2, robust enough for occasional high-bandwidth or low-latency data, and smart enough to automatically adjust the emissions posture to minimize the probability of counterdetection while maintaining adequate connectivity. AI should manage the network with a flexible hierarchy of redundant links.
Instead of a “card of the day” with fixed frequencies and call signs, the convoy’s collective AI would continuously negotiate a dynamic communications matrix, with circuit and cryptography kick procedures based on threat and environmental factors. AI also would evaluate and assign message precedence to ensure timely delivery. AI with machine-learning algorithms would need to be trained with copious and well-groomed data for these and other tasks. Like human intelligence, AI learns and improves by churning through extensive but reasonably well-structured experiences.
During normal transit in decent weather, the AI would default to a free-space optical (FSO) system to link platforms with clear line of sight, a technology a team from Johns Hopkins University demonstrated at sea in 2017. A modulated mirror galvanometer laser system could continually re-aim the FSO beam, allowing a single system to maintain links with every ship in sight. The same attenuation that makes FSO systems so hard to detect, however, also makes them vulnerable to weather, so they would require well-camouflaged radio-frequency (RF) backup links.
Effective RF camouflage starts with a thorough knowledge of the local EM environment and patterns of life, such as common waveforms as a function of time and season. Software-defined radios (SDRs) on each platform, managed by cooperative AI, would collect and aggregate then mimic emitters common to each environment. These SDRs also would use pseudonoise waveforms and multi-band frequency-hopping to achieve reliability and low probability of detection.7 The backup links would balance link quality and probability of detection by preferring more-attenuated bands. They also should employ automatic level control at the transmitter as a function of the signal excess at the receiver.8
Leveraging the Network
AI should do more than simply maintain the network. It also should provide a powerful and adaptive tactical decision aid. An average chess player facing Garry Kasparov could do worse than consulting IBM’s Deep Blue AI before each move; a sleep-deprived tactical action officer facing a saturation ASCM attack could do worse than executing with a single click the tactical AI’s recommended course of action—sometimes called “brilliant C2”—for the convoy. Even so, AI is no silver bullet: It must be trained, exercised, and refined in peacetime if it is to be trusted in combat.9 And it needs timely, relevant, and well-placed sensor inputs to render effective tactical decisions.
Because there will continue to be too few manned escorts to screen convoys, unmanned platforms will be essential. These should be distributed to see over the horizon, listen beneath the thermocline, and even take opportunistic shots. A convoy should include several medium-displacement unmanned surface vessels (MDUSVs)—the Sea Hunter and its successors—with Towed Airborne Lift of Naval Systems (TALONS) as pickets, cooperatively employing radar and towed-array sonar. High-endurance unmanned aerial vehicles (UAVs) could provide continuous early warning. Even small escort vessels, lacking helicopter decks but equipped with stern boat ramps, might support this vital mission with amphibious UAVs, such as the Singular Flyox.
Escorts would not be the only unmanned vessels. Many of the vessels under escort should be configured as unmanned, semiautonomous vessels, similar to the “expedient leader-follower” model being explored for Army ground logistics.10 A distributed navigation AI, developed from the Sea Hunter program, could cooperatively maintain formation, with the escort’s tactical AI directing evasive routing.11
Traditionally, escorts have carried most of a convoy’s sensors because of slow networks, limited tactical C2, limited sensor availability, and the need for specialized operators. But new technology has reduced these limitations. Every platform in the convoy—including merchants—should carry electro-optic (EO) sensors and a basic electronic warfare (EW) suite. Distributed EW suites integrated into the software-defined radios would speed detection and localization of ASCM radars and other emitters, which the AI-controlled EW network would then cooperatively jam, spoof, or mislead in conjunction with chaff and decoys.12 Each SDR could also function as a node in a multistatic “symbiotic” radar, which uses ambient signals (that is, not specifically radar emissions) to increase radar capability while reducing or eliminating the probability of detection.13 Machine-learning algorithms could help this cognitive RF network improve continuously.14 EO sensors, meanwhile, would add precision to detected tracks. Manned and unmanned escorts also would carry radars and active towed-array sonars.
From Track to Attack
Without capable and well-placed weapons, all of the above means little. The convoyed ships themselves could carry larger and better distributed magazines than their escorts.15 A four-cell vertical-launching system module fits nicely in a triple stack of 20-foot shipping containers, allowing a modest convoy to wield the same types of missiles as a surface action group, but in greater quantities and distributed for greater survivability: Evolved Seasparrow Missiles (ESSMs), for air and ASCM defense; Common Very Light Weight Torpedoes (CVLWT), atop ESSM or SM-2 boosters, for defense against submarine threats at most any range.16 Decoys, chaff, and directed-energy weapons distributed through the convoy and directed by tactical AI also could enhance a convoy’s self-defense capabilities.
The threat to joint maritime logistics and convoy operations has returned, but the next war at sea may not grant as much time to get it right as the Navy had in World War II. Each individual component of a winning technological solution may appear a panacea, but each can also fuel deadly overconfidence. To prevent this, the Navy must acknowledge the problem, then relentlessly test, integrate, and apply AI technology and operational concepts.
1. Quoted in Anthony W. Gray Jr., “Joint Logistics in the Pacific Theater” in Alan Gropman, ed., The Big ‘L’ American Logistics in World War II (Washington, DC: National Defense University, 1997).
2. Dennis M. Gormley, Andrew S. Erickson, and Jingdong Yuan, A Low-Visibility Force Multiplier: Assessing China’s Cruise Missile Ambitions (Washington, DC: National Defense University, 2014), 4.
3. Gormley et al., 16; Navy Recognition, “Submarine-Launched Variant of China’s YJ-18 Supersonic Anti-Ship Missile Emerges,” Navyrecognition.com, 2 October 2017.
4. David B. Larter, “‘You’re on Your Own’: US Sealift Can’t Count on Navy Escorts in the Next Big War,” Defense News, 10 October 2018.
5. “AI, Radiology, and the Future of Work,” The Economist, 7 June 2018.
6. John Wallace, “Johns Hopkins Team Demonstrates High-Bandwidth Free-Space Optical Communications at Sea,” Laser Focus World, 28 August 2017.
7. Colin Brown and Philip J. Vigneron, “Adaptive Use of Spectrum in Frequency Hopping Multi-Band Transmission,” NATO (2006); Tony Kendall, “The Knowing Sea: The Dawn of Smart and Ubiquitous SDR,” U.S. Naval Institute Blog, 15 January 2019.
8. Boulat A. Bash, et al., “Hiding Information in Noise: Fundamental Limits of Covert Wireless Communication,” IEEE Communications Magazine 53, no. 12 (December 2015).
9. Kimberly Underwood, “Artificial Intelligence Use in Command and Control,” AFCEA, 18 May 2018.
10. Uriel Epshtein and Charles Faint, “That’s Logistics: The Autonomous Future of the Army’s Battlefield Supply Chain,” Modern War Institute, 15 January 2018.
11. Elee Wakim, “Sealift Is America’s Achilles Heel in the Age of Great Power Competition” War on the Rocks, 18 January 2019.
12. For an analogous application, see Mark J. Mears, “Cooperative Electronic Attack Using Unmanned Air Vehicles,” Air Force Research Laboratory/Air Vehicles Directorate (2006).
13. See Amit K. Mishra and Michael Inggs, “White Space Symbiotic Radar: A New Scheme for Coexistence of Radio Communications and Radar,” 2015 IEEE Radar Conference, Johannesburg, South Africa (October 2015), and Kristine Bell, et al., “Cognitive Radars: A Reality?” (February 2018) arxiv.org/ftp/arxiv/papers/1803/1803.01000.pdf.
14. Felix Smits, Albert Huizing, Wim van Rossum, and Peter Hiemstra, “A Cognitive Radar Network: Architecture and Application to Multiplatform Radar Management,” Proceedings of the 5th European Radar Conference (October 2008).
15. CAPT R. Robinson Harris, USN (Ret.); Andrew Kerr; Kenneth Adams; Christopher Abt; Michael Venn; and COL T. X. Hammes, USMC (Ret.), “Converting Merchant Ships to Missile Ships for the Win,” U.S. Naval Institute Proceedings 145, no. 1 (January 2019).
16. Bryan Clark, Commanding the Seas: The U.S. Navy and the Future of Surface Warfare, Center for Strategic and Budgetary Assessments (2017), 25–26.