Scott D. Orr
In the modern era, rapid advances in science and engineering—most strikingly the leaps and bounds seen in World War II—have compressed the cycle of technological innovation in warfare to a pace much quicker than those of earlier periods, which saw decades or even centuries between major developments. Nonetheless, given the glacial rate of military procurement since the late 20th century,[1] deploying entirely new systems during a war that lasts less than several years is normally out of the question. Indeed, the hypothetical high-intensity great-power conflicts contemplated in recent decades could be described as “come-as-you-are”: enormous losses over a period of months or even weeks, combined with increasingly long production times, would make it difficult to replace equipment during the war or even to supply spare parts, let alone to deploy anything new. Before 2022, such a scenario was a matter of speculation, but we may now be witnessing such a war in Ukraine: by one estimate, Russia lost a third of its invasion force in the first 80 days of fighting, while the combined manufacturing capacity of the U.S. and its allies may take years to replace the munitions already expended by Ukraine’s forces. Three months into the war, far from deploying newly developed weapons, Russia was pulling half-century-old T-62 tanks out of storage.
And yet the war in Ukraine also provides an early test case of a new and very different trend in warfare, in the form of Ukraine’s Aerorozvidka, a nimble, highly adaptable unit fielding reconnaissance and strike drones. Aerorozvidka has showcased two new developments—systems defined by software, and field-swappable modular components—that will make technological innovation part of operational and even tactical decision-making. As militaries field customizable devices, we will move from “come-as-you-are” warfare to “fast fashion”: like fast-moving clothing brands, opposing commanders will compete to create systems that respond to new battlefield realities in a matter of days—or sooner.
What will this new state of military affairs look like? We don’t know—we’ve never seen anything like it before. Given the enormous uncertainty, we should allow clever, creative leaders to take the first crack at providing answers, by competing against one another in wargames that seek to map this vast new decision space. Their bright ideas will pave the way for rigorous analysis.
Would You Like to Play a Game?
The combat experiences of commanders and soldiers have always helped to drive and shape technological innovation, but once battle has been joined, a military force has had a certain set of tools at its disposal, with little or no means for creating new ones. There is no precedent for a world where routine construction of new tools by soldiers in the field becomes an integral part of warfare; we don't know what commanders might attempt, and so we cannot analyze their choices. There is no empirical evidence that operations researchers can use to build simulations, and no real-world experience to inform the judgments of military strategists.
As a data scientist, I appreciate the profound consequences of this shift in the nature of warfare. As a wargame designer, I believe we can start to come to grips with the shift by using wargames to identify commanders’ options for creating innovations on the battlefield. In most cases, wargames do not constitute rigorous analysis, but good wargames can spur creativity by forcing participants motivated to win to make critical decisions in the face of constraints—and in competition with live opponents—and these decisions will provide a basis for later analysis.
What’s New Here?
I use the term “software-defined systems” to refer to any military systems whose performance depend in great part on software that can easily be updated in the field. Software-defined systems, especially those driven by artificial intelligence (AI) and machine learning, make it possible to alter the parameters of military systems in the field literally at the touch of a download button. I use “modular systems” as shorthand for systems built from modular components that can easily be swapped in the field without specialized infrastructure. Modular components allow a single platform to perform very different roles (e.g., reconnaissance, attack, communications) at different times by plugging in different modules. By combining modular parts and modified software, an operator can in effect create new weapons systems from scratch, on the spot.
Neither type of system is entirely new: the F14A Tomcat, first flown in 1970, carried a programmable Central Air Data Computer and military aircraft have carried swappable payloads since well before World War II (and not just weapons—most of that grainy black-and-white footage of smoking warplanes spiraling to earth was taken by cameras installed in machinegun mounts). However, a variety of factors have converged to greatly increase the importance of software and modular components.
On the software side, two trends have driven this increasing importance. First, software’s role in many systems has grown over time, to the point that software has taken over functions formerly performed by hardware, as seen in software-defined radios (SDR’s). Second, due to a number of innovations such as Agile, DevOps, and containerization, the speed of the software development cycle has greatly increased. Moreover, for systems controlled or partly controlled by AI, when feedback from the field suggests the needs for changes, a model can be trained on new data, or on old data using different “hyperparameters”, thereby altering a system’s performance characteristics without requiring any new work by computer programmers.
Improvements in software development—particularly the adoption of common standards for interoperability between sub-systems (such as weapons and sensor packages) and for networking separate systems—have also greatly enhanced the battlefield potential of modular systems. Common standards for the hardware itself, especially universal interfaces (think of USB), have made it easier to plug together the physical parts. Modular systems may take the form both of platforms that can switch out multiple modules and combine them in novel ways, and of “systems” composed of multiple interoperating platforms—from “swarms” of similar platforms to ad hoc collections of heterogeneous ones.
An Increasingly-real-world Example
Imagine a swarm of a dozen small unmanned aircraft—the same size as the drones operated by hobbyists[2]—carrying general-purpose computers (a Raspberry Pi weighs only a few ounces) and modular sensor packages. While an individual drone of this size can carry only a small payload, an operator could outfit each such platform with a different sensor and then knit them into a single composite system using software. For example, an algorithm might combine the input of a drone carrying a video camera with that of another carrying a laser rangefinder to create a 3D map of a target area; processing could even be offloaded to the computers carried by other drones in the swarm, allowing results to be rendered much more quickly, and minimizing the amount of data that must be sent back to the swarm’s controllers. The swarm might even be used as a “mini-cloud” (literally in the air) to provide processing power to nearby ground units, especially light or dismounted infantry, who can’t lug around all the equipment that mechanized troops can. This “edge computing” would make the swarm resistant both to active electronic warfare (jamming), which struggles to suppress short-ranged communications, and to passive electronic warfare, since close-in low-power transmissions are difficult to detect.
Networked platforms could also create kinetic effects. Imagine the same swarm of small drones linked to a somewhat larger partner[3], big enough to loft the 15 kg payload of a light stand-off weapon like the U.S. AGR-20, a laser-guided rocket that can be carried by a wide variety of aircraft. One of the smaller drones could provide close-in sensors, while the others relayed targeting data through a chain of low-power directional signals to the weapon platform, which could then launch a non-line-of-sight attack from a protected position concealed by trees or bushes. Because the entire system would be networked, the armed drone would be able to use the spotter drones’ sensors as if they were its own, responding to input from the entire swarm in real time. With only one tiny drone near the target, the enemy might find it hard even to detect the attackers; countering the entire composite weapon system would prove very difficult, especially since a spotter drone, if destroyed or disabled, could quickly be replaced by another.
The first draft of this article, written in late 2021, relied entirely on speculation. Now, however, we have seen an early version of this type of warfare on the battlefield in Ukraine, in the form of Aerorozvidka, a Ukrainian unit formed by drone hobbyists and sustained by crowdfunding, which has used quickly modified drones to conduct reconnaissance shared through a force-wide intelligence network, and to carry out selective attacks on key vehicles, helping to stop the infamous 40-mile (65 km) convoy approaching Kyiv from the northwest during the early days of the Russian invasion. Even this nascent and under-resourced effort represents a significantly more advanced capability than the much-heralded use of artillery targeting drones by Russian forces in 2014 and 2015 in Ukraine. In that case, drone operators relayed information through human communication channels to artillery crews, resulting in delays of 10-15 minutes between drone arrival in the target area and the impact of indirect fire.
While studying the exploits of Aerorozvidka is fascinating, when both sides in a conflict begin to employ similar capabilities, the true analytical problem will arise. During combat, operational and tactical leaders will have a range of options that have never existed before. Indeed, given the number of levers that a commander could pull, in terms of software, AI models, platform configuration, and combinations of interoperating platforms, the decision space will expand enormously—and when the many new options available to one commander are multiplied by the options available to his or her enemy counterpart, the number of possibilities in the decision space will (literally) increase exponentially.
For instance, in the scenario above, of the networked weapon platform firing from cover, the opposing force will have its own options for countering the threat: Maybe a swarm of tiny drones assigned to track and target enemy spotters? Or maybe a localized electronic countermeasure (ECM) constructed from a software-defined radio (or radios), to jam the low-power transmissions? And this handful of examples notwithstanding, we don’t even know what all the possible options are. When opposing commanders can order up new systems with a time horizon that rivals Amazon Prime, technological innovation will become a part of the observe-orient-decide-act (OODA) loop, spurring an unimaginable variety of careful planning and creative improvisation at both operational and tactical levels. The world’s militaries have never encountered anything like this before, and the resulting lack of experience and empirical evidence will make effective analysis of this new era of warfare extremely problematic.
Wargaming to Map the Decision Space
Wargames can bridge this gap. Despite their long pedigree, opinions differ on the proper use of wargames by military planners, especially when it comes to their role in rigorous analysis (some tout the potential of wargaming as a method of analysis in itself, while others see it more as a complement to the analysis of operations researchers and military strategists). What everyone agrees on, however, is that wargaming excels at generating new ideas: surprisingly often, players in a game try things that have never occurred to anyone before, including themselves. Wargames are never fully accurate simulations, and the idiosyncrasies of each player and play-through make scientific reproducibility impossible—but that’s the point, to discover the unexpected possibilities that would be ironed out in a controlled experiment suitable for statistical analysis.
Specifically, the heat of competition between opposing teams—the desire to do everything possible to win—leads participants to come up with ideas that might not come to mind in the cool light of rigorous analysis. Where the key subject of interest is how commanders will respond to one another’s decisions, and to those responses in turn, on and on in a long chain of interaction, wargaming becomes even more revelatory: as hard as it is for an analyst to predict what a commander might do in the heat of combat, trying to get inside the heads of two people vying to best each other is an order of magnitude harder. However, the decisions made by wargame participants will provide the starting point for rigorous analysis. Once the decision space has been mapped—once we have a decent idea of all the moves and countermoves that might be made—operations researchers and other analysts will have something meaningful to work with.
Of course, reaching this objective requires an effective wargame design. As many of the sources cited above have noted, a “BOGSAT” (“bunch of guys/gals sitting around a table”), where participants meet face-to-face and simply discuss how their real-world counterparts would handle the scenario at hand, accomplishes little, typically becoming a forum for debate rather than an arena for experimentation. (One other consequence is that a handful of strong voices tend to dominate discussion, thereby limiting the number of new ideas put forward.) An effective wargame places players into the roles of real-world counterparts—fighting it out over a map rather than talking it out over a table.
Moreover, even with the best wargame design, actually executing the design effectively can prove difficult. Finding the right wargaming professionals is hard enough, but it can be even harder to keep a game on track in the face of all the institutional pressures that arise when interested parties make enthusiastic but misguided attempts to steer the process. And ensuring that operations researchers and other analysts make proper use of a wargame’s findings poses yet another hurdle.
Nonetheless, conventional analysis, while much more straightforward, would prove even more problematic: the terrain that needs to be explored in the case of software-defined and modular systems is so novel that wargaming is needed to show the path forward. In addition, wargames that include many different participants would serve a secondary purpose, of familiarizing officers with the new technologies and making them more comfortable to employ those capabilities in combat.
Getting from Here to There
What, then will an effective wargame on software-defined and modular systems look like? The considerations outlined above suggest a research program based on the following principles:The players will take on realistic roles, as commanders and staff officers, so that they face the same choices their real-world counterparts would; they will occupy those roles on adversarial teams, to encourage the competitive drive that inspires creative problem-solving.
The players will make their decisions bound by rules that represent a passable representation of reality—not because we hope to predict outcomes accurately, but because only a realistic decision process will produce realistic decisions, and the goal of this exercise is to document the decisions (and perhaps the decision process), not the outcomes. Cooperation between wargame designers and subject-matter experts (SME’s) on the new technologies (e.g., software developers, data scientists, and engineers) will be critical, as this is not a job that can be accomplished through a quick round of internet research. Instead, it will require an iterative back-and-forth between game designers and SME's on what is and isn’t possible, and how things should be represented in the game.
The rules will be clear and cover as many situations as possible, in order to focus players on making choices under constraints, rather than lobbying the umpires. During the early stages of the program, rules will change quickly as the designers—helped by players and SME’s—work out the best ways to represent reality, and add newly invented tactical options, but the idea is to improve the rules from one iteration to the next, not to throw them out mid-game in favor of subjective judgments. There will certainly be situations that the rules do not anticipate, and this is why wargames have umpires in the first place, but the fewer subjective decisions the umpires have to make, the more the players can concentrate on their own decision-making. What’s more, a clear, comprehensive set of rules makes computerization possible (see below).
To maximize the generation of ideas, a game will be run multiple times, with different players, because players with different personalities and backgrounds will come up with different ideas and use different approaches to decision-making—and indeed, the same players will often make different choices against different opponents. Note that this approach is a bit unusual by wargaming standards—it’s not unusual to run a game multiple times with slightly different scenarios, to test out different situations, but it is unusual to run the same game multiple times with different players. Running a game multiple times is costly, but this costliness has arguably produced an ingrained habit among wargamers of rarely even questioning whether multiple runs are needed. Given the idiosyncrasies of individual players, it’s never a bad idea to run the same game with different players when resources are available—and in this case, given the novelty of the domain, getting input from many different minds is crucial, making the extra expenditure of resources very much worthwhile.
Playing games on computers can enable the participation of as many players as possible, from different backgrounds and locations. A flexible platform such as the Standard Wargame Integration Facilitation Toolkit (SWIFT) can minimize the time required to develop a computerized version of a game. However, the programming labor required to code new rules into such a platform is not insubstantial, and therefore computerization makes sense only for the later stages of the program, once the rules are fairly well settled.
All of this will be administered by wargaming professionals experienced in researching, designing, organizing, and refereeing wargames—and empowered to maintain the integrity of the exercise, preventing a game from being derailed by participants intent on validating pre-conceived notions rather than exploring new ideas, or participants who are simply a little too intent on winning to accept the rules given to them.
Admittedly, such a research program sounds quite ambitious, but since software-defined and modular systems have barely been used in combat at this point, a force that wants to fight effectively has no other good options for defining the decision space that must be analyzed. The alternative is to stumble into a war unprepared, and then analyze the empirical evidence after the fact—or hope that someone else’s military is dumb enough to do so first.
No comments:
Post a Comment