Testing defense systems in operationally realistic scenarios is typically logistically difficult and expensive. For this reason, Bayesian methods have gained significant interest in recent years as a means of shifting testing "left" in the acquisition lifecycle-that is, integrating information from earlier phases of test to reach conclusions about system performance more quickly and to better infer operational performance when data from such scenarios is limited. Bayesian inference mathematically quantifies assumptions in the form of selecting prior distributions on the unknown parameters and strategies for integrating data collected under different conditions. In this article, we compare several Bayesian approaches for integrated test and evaluation, using the example of estimating the reliability of the Stryker family of vehicles from developmental and operational test data. We compute posterior reliability estimates for each method and conduct a sensitivity analysis to measure how each assumption influences the results. Altogether, the analysis not only shows the promise of Bayesian integration of information, but also the importance of careful and justifiable assumptions to ensure defensible results.
Surprise is inevitable in future armed conflict as adversaries innovate and evolve. Current exercises provide inadequate opportunities for combat support forces to develop necessary improvisational skills. This research proposes an extensible framework for human-in-the-loop control of exercises using a graph network for modeling combined with topological analysis and the resource-constrained project scheduling problem. This framework was assessed using the United States Air Force Silver Flag exercise. The performance of 10 solvers was used to show how users can trade optimal solutions for faster, good-enough solutions with results in 40% to 85% shorter times while producing substantially similar exercise schedules. The case study demonstrates the utility of a tool for structuring investigation and performing real-time analysis to maximize training value.
We study the mission planning problem of an unmanned aerial vehicle (UAV) operating in a hostile environment. The mission aims to neutralize the targets, which requires decisions for selecting targets, determining the target visitation sequence, loading payloads, allocating payloads to the selected targets, and specifying delivery locations and distances for payload-target engagements. Each target holds importance for the mission, and the objective is to maximize the overall utility collected from the entire terrain by neutralizing targets under capacity, time, and safety restrictions. We first develop approaches to represent the movements of the UAV and payload delivery in continuous space. Our model allows the UAV to approach targets from one of the multiple waypoints and release payloads from a continuum of options within the distance between the approached waypoint and the target location. We then develop a mathematical optimization model, which is linearized into a mixed-integer programming model, enabling us to derive optimal decisions for the routing and allocation requirements of the problem. We perform a comprehensive computational study to assess the impact of problem parameters and to identify model limitations. The results highlight the impacts of constraints, the ability of the model to handle varying levels of complexity, and applicability of the model to real-world scenarios(.)
This study examined organizational and human factors influencing the leakage of military secrets in the Republic of Korea (ROK) Armed Forces. Insider threats, highlighted by various military secret leak incidents, remain as crucial as external cyber-attacks. Through focus group interviews (FGIs) and analytic hierarchy process (AHP), the study explored how defense mechanisms, organizational commitment, and morality affect decisions to leak secrets. Key defense mechanisms (rationalization, projection, identifica-tion, denial) reveal how insiders justify misconduct; rationalization emerged as most influential. The study also investigated affective, continuance, and normative organizational commitment, uncovering how emotional bonds operate within the ROK military's hierarchical, collectivistic culture. It further delved into the significance of morality, particularly moral standards, motivation, and self-control, highlighting that robust ethics curb unethical acts. Drawing on theories by Dan Ariely, Norbert Schwarz, and Jonathan Haidt, the research employed Edgar Schein's Iceberg Model to assess cultural factors-artifacts, shared values, underlying assumptions-that shape insider behavior and leak risks. Grounded in these findings, the study recommends programs addressing self-rationalization, reinforcing moral values, and boosting organizational commitment. Though centered on the Korean military, its insights can guide security strategies globally, integrating organizational culture and human behavior for more effective protection of secrets. It provides a blueprint for mitigating insider threats.
The distribution of theater defenses involves strategically deploying weapons to safeguard targets against offensive attacks and enhance defensive effectiveness. This is a critical military concern, particularly evident in Vietnam. This study delves into two defense issues: the anti-aircraft launching assignment problem, which calculates missile deployment against attacking aircraft to optimize defense, and the anti-aircraft mission planning problem, which determines optimal battalion locations alongside deployment assignments. Both problems are NP-hard, addressed through exact and heuristic solution approaches. Mixed integer linear programs were developed for optimal resolution via the Cplex solver, while tabu searches offered metaheuristic solutions. Exact methods suit offline scenarios, while metaheuristic approaches are more fitting for online scenarios. Comprehensive experiments validate the efficacy of both methods. Subsequently, a tool integrating the proposed algorithms was developed, tailored for educational use at an air defense-air force academy in Vietnam.
We establish a framework that links time-limited search problems with prize-collection problems. The searcher's objective is to travel a path with a particular origin and destination so that the probability of finding the entity within the available time budget is maximized. As the searcher traverses the network, they learn more about their surroundings. Hence, the a priori probabilities need to be updated so that the searcher can choose the next moves using real-time data. Other variants of the base search problem are also studied, including searching for two immobile entities with (in)dependent locations and searching for an unknown number of entities. Our theoretical analysis demonstrates that learning does not occur in the base search problem. Thus, it is the same as the prize-collection problem and can be solved to optimality. Conversely, after mapping the other extensions to the prize-collection problem, finding the optimal solution is not guaranteed due to the presence of learning. Computational experiments as well as a military and security case study are presented at the end, the latter taking practical considerations, including uncertainties in traveling times and collaboration between multiple searchers into account.
Double counting, or data incest, is a data fusion problem that arises in networks when members (sensors) of a decentralized network are tasked to make estimates based on shared signals. Unknowingly counting the same signals multiple times, due to an inability to keep track of the peer-to-peer signal propagation paths, the network members obtain wrong estimates of the environment, while unnecessarily overusing the network's time and energy resources. Indeed, energy spent to propagate a signal is at a premium for many real sensor networks, especially in passive surveillance environments, e.g., at sea or in the Arctic. This article proposes a practical communication protocol that avoids double counting by using networks of a topology known as cactus graph topology. We present exact and heuristic cactus network design methods to minimize energy expenditure in disseminating signals. In addition to centralized graph structure traversal algorithms, a decentralized algorithm is presented that is a self-organizing, lightweight cactus network design approach that any placed sensors can adopt at the stage of establishing network-wide communication. The suboptimal, low-computational-expense solutions found with the developed heuristic and decentralized algorithms compare well to the optimal solutions on several test problem instances.
The study aims to determine the Air Force University military pilot-cadets' predisposition toward risk taking and propose specific recommendations to enhance their professional training based on the findings. This qualitative study was conducted in September 2021 and May 2022 at the Ivan Kozhedub National University of the Air Force. Schubert's "PSK" technique was employed for the diagnostic study, effectively evaluating human behavioral responses in life-threatening situations, including extreme conditions. The study demonstrated that most military pilot-cadets who underwent diagnostics exhibited satisfactory readiness for professional risk. This finding supports the importance of tailored training programs that prepare cadets for various risk scenarios during combat operations. By addressing the multifaceted nature of risk-taking predispositions in military aviation, educational institutions can foster an optimal balance between caution and decisiveness, ultimately enhancing the effectiveness and safety of future combat air operations. Military pilot-cadets' professional training focuses on developing their cognitive readiness to process complex information quickly. This aspect is crucial in executing combat missions and minimizing pilot risk. This study optimizes aviation branch allocation by assessing cadets' risk taking, enhancing readiness.
Aviation safety in the United States (U.S.) military has received growing attention in recent years due to numerous high-profile mishaps. Despite the increased attention, there have been few quantitative analyses of the relationship between pilot attributes and mishap rates. In this study, we use nearly 15 years of U.S. Air Force (USAF) safety and administrative records to investigate the relationship between pilot attributes and fighter aviation mishap rates. First, we present an analysis of flight mishap rates for different mishap classes and fighter aircraft types, referred to as a mission design series (MDS). Second, we quantify pilot attributes and present an analysis of fighter pilot populations across time and MDS. We then model the association between pilot attributes and annual rate of class A, B, and C flight mishaps, which we refer to as high-class mishaps (HCMs), using a Bayesian regression framework. Our results show prior flight experience and key characteristics of an MDS pilot community are associated with the rate of HCMs. Specifically, we find that MDS pilot communities with 10 more flight hours in the past year are, on average, associated with a 5% decrease in HCM rate. Additionally, we find that a 0.1 standard deviation increase in the proportion of pilots who are instructor pilots, distinguished graduates from commissioning source, and graduate degree recipients is associated with a reduction in major aviation mishaps by 2.1%, 2.0%, and 1.3%, respectively. These find-ings have significant financial implications, given that the cost of a single HCM starts at $50K and can be as high as $200M. In addition to our model results, our efforts to quantify pilot attributes and model the relationship between personnel factors and mishap rates using Bayesian regression and predictive projection for feature selection represent a valuable methodological contribution to aviation accident analysis.
Soldiers at military schools in the United States Army frequently take part in simulated platoon-level operations where graduation depends on receiving a positive evaluation in specific leadership roles. Scheduling leadership role evaluations is a nontrivial task due to the numerous constraints placed on the scheduler, including but not limited to balancing mission difficulty, maintaining unit organization, and avoiding consecutive evaluations. Scheduling leadership roles currently relies on manual allocation by instructors, typically without a systematic approach. A nonsystematic approach has the potential to produce biased and unfair schedules that are marked by the inability to adapt the schedule after evaluation failures or unexpected events such as injuries or illnesses. We develop an integer linear programming model that efficiently and fairly distributes leadership roles among trainees for the entirety of a training exercise. The model we present minimizes the time needed for every trainee to receive their required evaluations, which in turn allows an instructor maximum flexibility to alter the evaluation schedule as needed, without needing to consider potential negative impacts on any trainee's graduation chances.
This study addresses a demand forecasting problem for military spare parts with unstructured historical data. The problem is to determine the procurement while satisfying spare parts demands and budget constraints in each period of a planning horizon. As an extension of previous studies, initial provisioning (IP) spare parts are considered with the objective of maximizing demand accuracy. IP parts are necessary to perform a given mission during the initial period of acquisition of a weapon system. IP parts are often subject to separate demand forecasting and inventory management processes from other types of spare parts. To solve the demand forecasting problem with the characteristics of IP, machine learning with the Backward Elimination method is proposed. Computational results show that the machine learning-based approach outperforms the others.
Maximizing military force efficiency requires tasking weapon systems across all echelons, domains, and services to achieve desired target effects. Early research on flexible kill chains led to the Information Age Combat Model (IACM), which uses a graph-analytic approach to represent connections in the battlespace. These principles are relevant to emerging concepts like Joint All Domain Command and Control (JADC2), enabling kill paths irrespective of branch and service alignment. However, the original metrics for evaluating kill chains are computationally intractable. This article proposes a new algorithm that enumerates and calculates metrics for all kill paths. Implemented as a computational decision support tool, this algorithm detects and enumerates all kill paths, calculates values for attack vectors, and allocates influencers to targets both statically and dynamically as new targets are sensed. It also identifies critical components requiring special protection. The algorithm operates efficiently enough for use within simulations, providing a basis for command-and-control decision making. Its feasibility, applicability, and computational efficiency are demonstrated through selected simulation runs. Although our demonstration utilizes IACM kill chains, the algorithm can be extended to various kill chain types and integrated into battle management systems for operational support.
Automation of jobs has been a prominent topic of investigation in recent years. As automation may relieve military personnel from tasks that are "dull, dirty, and dangerous," automation will be crucial to maintaining military advantage. However, as there are also security, legal, and ethical issues related to automation in a military context, the effect of automation on armed forces might differ from that on civilian organizations. Therefore, the impact of automation on military organizations should be studied specifically. This article investigates whether routine intensity of jobs, bottleneck skills for automation, and educational level affect automation in the Norwegian Armed Forces. We use a binary logistic regression model to analyze our survey data. We find that routine intensity of tasks led to automation among our respondents. However, the effect is small. Three of the bottleneck skills seemed to prevent automation among the respondents: collaboration, communication, and creativity. Civilian educational level had a large effect on automation. We found that military combat personnel experienced less automation than military noncombat personnel and civilians.
Some government and corporate decisions are hierarchical in two dimensions: a hierarchy of alternatives and a corresponding hierarchy of decision makers. An example of such a hierarchical decision process is the U.S. Department of Defense Program Objective Memorandum (POM), which sets development and acquisition plans within a given budget. At the top level of the hierarchy, senior leaders set directions for those acquisition and development plans, directions that can be viewed as or translated into families of portfolios called henceforth Programs. Programs comprise projects that are the eventual fundable entities. Although "hierarchy" is a core feature in this decision-making setup, it does not comply with the well-known analytic hierarchy process, where decision alternatives are at the bottom level of a hierarchy that also includes goals and criteria. In this article, we propose a modeling framework of a different type, where the hierarchy only comprises alternatives; the criteria, which may be alternatives dependent, are "orthogonal" to the levels of the hierarchy. We develop a methodology for handling such a decision setup and demonstrate its application in reference to the POM. The multicriteria-decision-analysis part of the methodology hinges on the widely used concept of least squares.
This research studies the optimal mix of theater airlift and sealift vehicles needed to conduct military sustainment operations. Theater aircraft deliver supplies quickly, but the payloads are smaller compared to sealift vessels. Conversely, theater sealift vessels deliver more supplies per mission, but are slower than aircraft. In this article, we construct a mixed integer program to identify the optimal fleet size and composition, or mix, of theater vehicles to deliver cargo from a single, central depot to theater outposts. Three operational factors are studied in this research: distance to the outpost, number of people at the outpost (with more people requiring more supplies), and amount of supplies available at the outpost at the start of sustainment operations. To explore the solution space, we construct several experimental designs with three operational factors, each with three levels. Next, we solve mixed integer programs and conduct regression analyses of the results to quantify the relative importance of the factors. This research provides new insights into the optimal mix of vehicles, which had previously not been empirically assessed in a military context. Military commanders could use the insights from this article to ensure enough theater vehicles, by type, are deployed to execute sustainment operations.
The array of specializations in a military force is chosen to deter potential adversaries and assure success in conflict when deterrence fails. However, effectiveness of military force in future conflict is often uncertain. Hence force planners must identify critical goals of future conflict and must maximize robustness against uncertainty in achieving these goals. The goals of conflict are not maximized: robustness against uncertainty is maximized and goals are satisficed. This methodology is called robust satisficing. Two propositions are proven, employing the concept of robust dominance. One force composition is robust dominant over another force composition if the first composition assures adequate military effectiveness over a wider range of uncertainty in the effectiveness. Proposition 1 establishes a sufficient condition for one force composition to be robust dominant over another force composition, if both compositions have the same predicted military effectiveness. Proposition 2 considers two force compositions whose estimated military effectiveness are not the same, and the composition with lower estimated effectiveness is less uncertain than the composition with greater estimated effectiveness. The proposition establishes that neither composition is robust dominant over the other and establishes ranges of critical effectiveness for which each composition is robust preferred.
