90 Seconds: Human Command Versus Machine-Speed Warfare

Full report

On this page

The Compression of the Kill Chain in Modern Combat

The contemporary battlespace is defined by the exponential proliferation of data, the hypersonic velocity of incoming threats, and the rapidly shrinking window for human cognitive processing. Historically, the military targeting process—often codified in doctrines such as Decide, Detect, Deliver, Assess (D3A) for land forces or Find, Fix, Track, Target, Engage, Assess (F2T2EA) for joint and air operations—relied on sequential human communication spanning minutes, hours, or even days1. These traditional linear kill chains depended on manual data correlation, hierarchical command approval, and physical cross-domain coordination1. Today, the integration of artificial intelligence, multi-sensor data fusion, and preprogrammed rule sets has compressed this sequence into seconds, transforming the linear kill chain into a highly dynamic, automated kill web. Public perception often incorrectly attributes this speed to a sentient, magical artificial intelligence that independently understands the nuances of war and conflict. In reality, automated defensive systems are highly complex, composite architectures that combine discrete hardware and software components. They do not comprehend conflict; they execute probabilistic mathematics, kinematic state estimations, and Boolean logic gates4. The central challenge of modern defense design, particularly within initiatives like the Department of Defense's Joint All-Domain Command and Control (JADC2) and the Defense Advanced Research Projects Agency's (DARPA) Mosaic Warfare concept, is not necessarily removing the human from the loop, but determining exactly when and where the human makes the decision to engage6. To elucidate this paradigm shift, the following report outlines the architectural design, technical mechanics, and real-world context for an interactive educational experience titled “90 Seconds: Human Command Versus Machine-Speed Warfare.” This simulation serves as an analytical tool to demonstrate how human agency is shifted backward in time, shaping machine-speed responses through pre-mission parameters.

System-Level Kill-Chain Architecture

To demystify the concept of autonomous weapons, the educational experience first presents a transparent, system-level architecture of the automated kill chain. This demonstrates how discrete algorithms and hardware execute the F2T2EA sequence in a machine-speed environment without requiring emergent sentience.

Find: Sensor Integration and Signal Processing

The sequence begins with heterogeneous sensors collecting raw environmental data, including radar radio frequency (RF) returns, infrared (IR) thermal signatures, acoustic waves, visual telemetry, and electronic intelligence (ELINT)8. In highly contested environments, adversarial platforms often feature low radar cross-sections or utilize electronic warfare to mask their presence. To overcome this, modern systems utilize algorithms such as Track-Before-Detect (TBD). Rather than waiting for a single strong radar return to cross a predetermined detection threshold, TBD algorithms accumulate weak signal energy over multiple frames, allowing the system to find targets that would otherwise remain hidden in background noise10.

Fix: Data-Fusion and Kinematic State Estimation

Once raw data is ingested, data-fusion systems must combine uncertain and asynchronous observations into a coherent spatial picture. Because different sensors have varying latency, processing speeds, and error rates, algorithms such as the Extended Kalman Filter (EKF) or Unscented Kalman Filter (UKF) are deployed5. The Kalman filter recursively predicts the target's kinematic state and updates the covariance matrix to account for uncertainty. The filter balances the system's predictive model against incoming noisy measurements using the Kalman gain equation: [Figure omitted from source export]5. In highly dynamic environments, advanced variants like the Fuzzy Adaptive Extended Kalman Filter (FA-EKF) dynamically adjust the observation noise covariance matrix to handle non-linearities and sensor degradation, mathematically correlating disparate sensor reports into a single, unified spatial track13.

Track: Predictive Modeling and Custody

Tracking involves maintaining continuous custody of the identified object as it traverses the battlespace. The kinematic tracking algorithms predict future positions based on velocity, acceleration, and maneuvering models14. If a primary sensor is temporarily blinded by weather or active jamming, the system relies on the predictive state estimation and secondary sensors (such as an optical camera confirming a radar track) to bridge the gap and maintain a continuous lock on the target's trajectory16.

Identify: Classification and Uncertainty Modeling

Automated systems do not output absolute certainty regarding a target's identity; they assign probabilities. Classification models utilize mathematical frameworks, such as Bayesian inference or the Dempster-Shafer theory of evidence, to fuse conflicting beliefs18. Unlike Bayesian methods that require mutually exclusive hypotheses, Dempster-Shafer allows the system to assign evidence to propositions and handle general uncertainty19. For example, an infrared sensor might suggest a high probability of a hostile drone, while a radar profile suggests a bird. The system fuses these basic probability assignments using the Dempster-Shafer combination rule to calculate a fused confidence score, transforming it via Pignistic probability into a measurable classification percentage18.

Prioritize: Weapon Target Assignment

When facing a saturation attack, the system must prioritize detected threats. This requires solving the Weapon Target Assignment (WTA) problem, a classic combinatorial optimization dilemma known to be NP-complete or NP-hard4. The WTA algorithm calculates the optimal pairing of available defensive interceptors to incoming threats to minimize the expected survival value of the attackers21. Because the computational complexity increases exponentially with the number of targets and weapons, the system relies on metaheuristics or branch-and-bound algorithms to generate near-optimal firing schedules in fractions of a second, accounting for interceptor kinematics, threat lethality, and available magazine depth23.

Authorize: Rule-Based Logic and Human Constraints

Once a firing solution is optimized, rule-based logic filters the allowable responses. This is the critical juncture where preprogrammed human intent is enforced. The logic engine checks the optimized target list against constraints established by commanders: geographic geofences, altitude floors, speed thresholds, operational time windows, and friend-or-foe (IFF) interrogations25. If all programmatic constraints are satisfied, the system either prompts a human operator for final authorization or executes the engagement automatically based on previously granted authority26.

Engage: Automated Effectors

Upon receiving authorization, automated effectors carry out the approved response. This involves transmitting the firing solution to kinetic interceptors (such as close-in weapon systems or surface-to-air missiles), electronic warfare emitters, or directed-energy weapons. The effector physicalizes the algorithmic decision, neutralizing the kinematic threat28.

Assess: Battle Damage Assessment

The final phase feeds information back into the system. High-resolution sensors and data-fusion loops evaluate the kinematic destruction, trajectory alteration, or behavioral cessation of the target. This real-time battle damage assessment allows the WTA algorithm to dynamically update, either cueing a re-engagement if the target survived or shifting interceptors to the next prioritized threat in the queue26.

Interactive Concept: Scene-by-Scene Narrative

To operationalize these concepts for an educational audience, the design includes a split-screen interactive simulation titled “90 Seconds: Human Command Versus Machine-Speed Warfare.” The simulation places the user in the Combat Information Center of a fictional guided-missile destroyer, the FNS Sentinel, tasked with defending a commercial convoy in the fictional Strait of Aethelgard. The environment is heavily congested with commercial maritime and air traffic, creating immense sensor clutter. The threat consists of a coordinated saturation attack: 12 autonomous fast-attack surface crafts and 24 low-observable loitering munitions deploying simultaneously from a disguised commercial container ship 15 nautical miles away.

Scene 1: The Pre-Mission Configuration Phase

Before the 90-second countdown begins, the simulation forces the user to establish the system's operating parameters, demonstrating that automation shifts human decision-making earlier in time. The user interfaces with a tactical map and a series of sliders. They must establish a geographic geofence, indicating that any object crossing a 10-nautical-mile perimeter is subject to evaluation. They must set a Dempster-Shafer confidence threshold, determining that the system requires an 85 percent probability of hostile classification before generating a firing solution. Finally, they define the Rules of Engagement (ROE) logic gates, programming the system to ignore any object squawking a civilian IFF code or traveling below 50 knots. The user commits these settings, and the simulation transitions to the live operational environment.

Scene 2: Mode 1 - Human-Directed Execution (Human-in-the-Loop)

The scenario begins. The user is in full manual control, required to review every sensor report and approve each action. The interface immediately floods with hundreds of raw, blinking radar returns and thermal anomalies. The user must manually click a radar track, attempt to correlate it with a separate optical camera feed, and request a classification. The human cognitive load spikes instantly. As the user struggles to identify the first three fast-attack crafts, the 24 loitering munitions penetrate the 10-mile geofence. The user attempts to calculate intercept geometries and assign weapons, but the sheer volume of data induces paralysis. At 42.5 seconds, the user manually authorizes the first interceptor launch. By 90 seconds, 21 hostile objects have penetrated the defensive perimeter, resulting in a catastrophic mission failure. The scene highlights the physical limits of human neurological processing in a machine-speed environment.

Scene 3: Mode 2 - AI-Assisted Execution (Human-on-the-Loop)

The scenario resets to zero. In this mode, the system's algorithms handle the Find, Fix, Track, Identify, and Prioritize phases. The user's screen is entirely different: raw noise is abstracted away, replaced by a clean graphical user interface showing unified tracks with calculated classification probabilities. The WTA algorithm presents a perfectly optimized firing schedule. However, the system requires a manual human button press to authorize every single engagement. The user must rapidly click "Authorize" 36 times as the system highlights targets in red. While the user successfully defends the convoy, the scene introduces the concept of the "irony of automation." The operator is reduced to a rapid button-clicker, experiencing profound automation bias. There is no time to verify the sensor data behind the AI's recommendations; the human is merely providing a false veneer of oversight, trusting the machine implicitly30.

Scene 4: Mode 3 - Preauthorized Automation (Human-out-of-the-Loop)

The scenario resets for the final time. The system is armed with the pre-mission parameters established in Scene 1 and granted full authority to execute the kill chain if all programmatic criteria are satisfied. The user is instructed to hold a physical "Abort" switch and monitor the engagement. Within 2.1 seconds of the threats crossing the geofence, the system detects, fuses, classifies, prioritizes, and engages all 36 targets. A flurry of interceptors is launched automatically. The swarm is entirely defeated in 18.4 seconds. The user watches the display, performing no actions other than maintaining situational awareness. The scene visually reinforces that while the machine acted autonomously in the moment, it was entirely bound by the human's pre-mission configuration.

Simulation Metrics, Branching Decisions, and Data Comparisons

To quantify the differences between the execution modes, the simulation generates end-of-action reports displaying structured data. These metrics prove that while automation vastly reduces elapsed time, it fundamentally alters the nature of human workload and target-classification uncertainty.

The data processing metrics highlight how algorithms shield the human from raw environmental noise, resolving conflicting observations mathematically rather than requiring human intuition.

The outcomes demonstrate the trade-offs between false-positive risks, operational success, and the origin of the engagement decision.

During the initial configuration phase, the user's choices create dynamic branching outcomes that govern the Mode 3 simulation. The following table outlines the user interactions and their cascading consequences on the automated system.

“What the Machine Saw” Replay Sequence

Following the completion of the three simulation modes, the user enters a diagnostic replay feature designed to dispel the illusion of AI sentience. The user can toggle between four distinct perspectives of a single loitering munition intercept, observing how data is transformed from physics to logic. The first perspective is the raw sensor feed. The screen displays static, thermal blooms, and erratic radar returns affected by multipath interference from the ocean surface. The user observes that the machine's actual vision is deeply flawed, chaotic, and heavily distorted by environmental physics. The second perspective reveals the fused AI model. The environment is stripped away, replaced by a wireframe grid overlaid with matrices. The user watches the Extended Kalman Filter in real-time. As an asynchronous radar ping and an infrared ping arrive, the error covariance matrix visually shrinks, collapsing the uncertainty cloud into a single, stabilized track vector8. Alongside the track, a continuously updating bar chart displays the Dempster-Shafer belief mass shifting from a 40 percent "Unknown" probability to an 89 percent "Hostile" probability as successive data frames accumulate18. The third perspective shifts to the human operator's Graphical User Interface (GUI). The complex mathematics of the Kalman filter and Dempster-Shafer matrices are entirely abstracted away. The interface simply shows a red diamond, a vector line, and a text tag reading "TRK 4011 \- CONF 89%." This view highlights the danger of the interface: the abstraction is necessary to prevent cognitive overload, but it simultaneously invites automation bias by presenting probabilistic guesses as absolute, clean facts30. The final perspective displays the final automated rule set. The visual interface disappears, replaced by lines of executing Boolean logic: IF (Track \== Hostile) AND (Confidence \> 85%) AND (Range \< 10nm) AND (IFF \== None) THEN (State \= AUTHORIZED); (Execute \= TRUE). This code execution proves that the machine did not decide to kill based on anger, tactical genius, or understanding of war; it merely satisfied a mathematical condition that was pre-authorized by the user's earlier inputs.

Real-World Context and Historical Precedents

To ground the interactive simulation in reality, the experience utilizes carefully sourced examples of defensive automation. The history of these systems reveals a persistent tension between the tactical necessity for machine speed and the cognitive limitations of human operators. The following case studies are structured to separate verified facts from interpretation and fictional elements.

1. The Aegis Combat System and USS Vincennes (1988)

Verified Public Facts: On July 3, 1988, the USS Vincennes, a guided-missile cruiser equipped with the advanced Aegis Combat System, mistakenly shot down Iran Air Flight 655, a civilian Airbus A300, killing all 290 people on board. The Aegis system's sensors and kinematic tracking properly recorded the aircraft's flight path, speed, and civilian transponder codes34. Analyst Interpretation: Investigations and subsequent human factors research highlighted phenomena such as scenario fulfillment and cognitive overload. Captain William C. Rogers III and his crew were operating under extreme stress, having recently engaged Iranian gunboats, and expected a hostile Iranian F-14 attack. Because the complex Aegis interface provided overwhelming amounts of data, the operators unconsciously interpreted the information to fit their predetermined scenario—believing the aircraft was descending into an attack profile when the system's raw data showed it was ascending35. Unknown or Classified Capabilities: The exact modern upgrades to the Aegis system's user interface and its classified algorithms for automated threat discrimination against modern hypersonic glide vehicles remain restricted. Fictional Simulation Elements: The FNS Sentinel in the interactive simulation operates a fictionalized interface inspired by Aegis, but simplified for civilian educational purposes.

2. Patriot Air and Missile Defense System (2003)

Verified Public Facts: During Operation Iraqi Freedom in 2003, US Army Patriot missile batteries were involved in two fratricide incidents, shooting down a British Tornado GR4 and a US Navy F/A-18 Hornet30. In both instances, the Patriot system misclassified the friendly aircraft as hostile tactical ballistic missiles or anti-radiation missiles and successfully engaged them30. Analyst Interpretation: A Defense Science Board report and Army Research Laboratory assessments pointed to undisciplined automation and severe automation bias. The Patriot system featured an auto-fire mode meant to defeat high-speed tactical ballistic missiles, which offered almost no human reaction time30. Operators were trained in an environment that fostered uncritical reliance on the machine, leading them to trust the system's algorithmic misclassifications over their own situational awareness30. Following the fratricides, operational procedures were altered to keep systems in manual or standby mode when dealing with certain threats, adding deliberate friction to the kill chain37. Unknown or Classified Capabilities: The current proprietary logic utilized by the Patriot system to filter false-positive radar returns in heavily contested electronic warfare environments is classified. Fictional Simulation Elements: The simulation's Mode 2 (AI-Assisted) replicates the psychological environment of the Patriot operators, forcing the user to rely entirely on the system's classification tags without time for independent verification.

3. Phalanx Close-In Weapon System (1991)

Verified Public Facts: During the 1991 Gulf War, the USS Jarrett was operating in the Persian Gulf near the battleship USS Missouri. When an incoming Iraqi Silkworm missile was detected, the Missouri fired SRBOC chaff decoys to confuse the missile's radar. The Jarrett’s Phalanx CIWS, operating in its fully automated target-acquisition mode, detected the chaff cloud, categorized it as a threat, and automatically engaged it, striking the Missouri with multiple 20mm rounds40. Analyst Interpretation: The Phalanx CIWS is widely cited as an early example of a supervised autonomous weapon. In its Auto-Fire mode, the radar searches, tracks, evaluates, and fires on incoming threats entirely without human intervention once activated43. The incident highlights the brittleness of automated logic in complex environments; the machine executed its programmatic instructions flawlessly, shooting at a radar return that matched its threat parameters, lacking the human contextual understanding that the return was friendly chaff. Manufacturer or Government Claims: Modern iterations of the Phalanx system claim significantly upgraded radar processing and electro-optical sensors to differentiate between threats and countermeasures. Fictional Simulation Elements: The simulation's Mode 3 utilizes the fundamental concept of the Phalanx's auto-fire mode, requiring the user to trust a pre-activated logic loop to defend the ship.

4. Raytheon Coyote and KuRFS Radar

Verified Public Facts: The US Army utilizes the Low, slow, small-unmanned aircraft Integrated Defeat System (LIDS), which pairs the Ku-band Radio Frequency Sensor (KuRFS) with the Raytheon Coyote interceptor45. The system is designed to detect and defeat drone swarms and has seen widespread operational use and combat intercepts in the Middle East46. Manufacturer or Government Claims: RTX (formerly Raytheon) claims the KuRFS radar provides persistent 360-degree detection and can identify objects as small as a 9mm bullet48. Furthermore, the Coyote Block 3 utilizes a non-kinetic warhead capable of autonomously defeating swarms of drones, minimizing collateral damage and allowing the interceptor to be recovered and reused48. Analyst Interpretation: The Coyote and KuRFS system represents the cutting edge of Joint All-Domain Command and Control (JADC2) sensor-to-shooter integration. As drone swarms increase in density, analysts note that systems like Coyote push the boundary of human cognitive capacity, necessitating higher degrees of automated target prioritization at the tactical edge to compress the kill chain26. Note: None of the aforementioned systems are portrayed as fully autonomous artificial general intelligences. They are strictly bound by their programmatic logic, algorithms, and human-defined operational modes. There is no reliable public evidence that any named system possesses independent, self-directed lethal autonomy outside of human-programmed boundaries.

Failure Modes and Cognitive Vulnerabilities

To ensure a nuanced educational experience, the simulation intentionally exposes the user to the inherent brittleness of machine-speed warfare. Automation does not eliminate error; it merely shifts the nature of the vulnerability from human physical limitations to systemic and cognitive blind spots. Adversarial spoofing presents a primary failure mode. Automated systems rely heavily on Kalman filters, which predict future states based on current kinematics. An adversary utilizing advanced electronic warfare can inject false radar returns—ghost tracks—into the receiver. If these spoofed tracks are designed to perfectly match the pre-authorized rule set (e.g., inbound trajectory, high velocity), the automated effector will engage the phantom targets, rapidly depleting the defender's magazine capacity and leaving the asset vulnerable to the actual kinetic strike. Sensor degradation introduces critical uncertainty. Heavy precipitation, sea spray, or smoke can severely degrade infrared and visual sensors. If a multi-sensor fusion engine loses a primary data stream, the Dempster-Shafer probability model may fail to reach the established confidence threshold. The system will default to an "Unknown" classification, forcing the human back into the decision loop precisely when the environmental conditions are most chaotic. Furthermore, human cognitive vulnerabilities persist even in highly automated environments. Research pioneered by Norman Mackworth demonstrates the "vigilance decrement," wherein human operators staring at radar screens experience a steep, biological drop in detection accuracy and situational awareness after roughly 30 minutes of monitoring rare events50. In the simulation, if the user operates in Mode 3 (Preauthorized Automation) and holds the Abort button during a prolonged period of inactivity, their reaction time to a sudden, erroneous AI targeting decision is artificially delayed to simulate this biological reality. This leads to the out-of-the-loop performance degradation described by Endsley's model of Situation Awareness (SA)52. SA requires perception (Level 1), comprehension (Level 2), and projection (Level 3\)54. When a system operates autonomously, the human loses Level 2 and Level 3 awareness. If the automation experiences a catastrophic failure and control reverts to the operator, the human is violently thrust back into the loop without the contextual understanding required to make a safe, legal, or tactically sound decision, often resulting in paralysis or fatal errors52.

Safeguards, Policy, and the Shift of Human Control

The integration of automation in lethal systems is heavily regulated by policy to mitigate these technological and cognitive failures. United States Department of Defense Directive (DoDD) 3000.09, Autonomy in Weapon Systems, provides the foundational framework. The directive mandates that all autonomous and semi-autonomous systems be designed to allow commanders and operators to "exercise appropriate levels of human judgment over the use of force"55. Crucially, DoDD 3000.09 does not universally mandate a "human in the loop" at the exact moment of a trigger pull for defensive systems57. The policy recognizes that for systems like Aegis missile defense, Phalanx CIWS, or counter-swarm capabilities, inserting a human into the final millisecond is physically impossible and tactically fatal. A strict human-in-the-loop requirement would negate the very speed advantage the system was designed to achieve. Instead, "appropriate levels of human judgment" are implemented through rigorous, systemic safeguards that shift human control leftward on the timeline:

1. System Design and Testing: Ensuring the algorithms perform predictably under realistic conditions, including verification and validation of the weapon system's safety rules and AI ethical principles56.

2. Human-Machine Interfaces (HMI): Designing interfaces that provide transparent feedback on system status to prevent automation bias, ensuring the operator understands exactly what the system is doing and why25.

3. Temporal Shift of Agency: The human exercises control before the engagement occurs. The operator sets the geographic geofence, establishes the Dempster-Shafer confidence thresholds, activates the system during a specific time window, and holds the ultimate authority to deactivate or abort the system's actions55.

The simulation visualizes this shift by emphasizing that the human's most critical, legally binding decisions occur at T-minus 60 minutes during the mission planning phase, not at T-minus 2 seconds during the kinetic intercept.

Glossary of Operational and Technical Terminology

Conclusion

The pursuit of machine-speed warfare is an operational necessity driven by the immutable physics of modern threats. As adversarial weapons cross the threshold from supersonic to hypersonic, and as autonomous drone swarms achieve unprecedented density, the human neurological system is simply too slow to execute the traditional F2T2EA or D3A kill chains. The transition to AI-assisted and preauthorized automated systems solves the kinematic problem of speed, enabling defenders to survive saturation attacks. However, this transition introduces profound cognitive vulnerabilities, including automation bias, vigilance decrement, and the loss of situational awareness. The central lesson of defensive automation is not that machines have replaced humans in the art of war. It is that the architecture of modern combat has fundamentally shifted the locus of human agency. A machine cannot understand the legal principle of proportionality, nor can it grasp the strategic ramifications of fratricide. It can only execute the probabilistic math, optimization algorithms, and Boolean logic embedded within it by its human creators. Therefore, the moral and legal responsibility for lethal action remains entirely human; it has simply been displaced backward in time and space. When a machine acts in two seconds, was the human removed from the decision—or was the human decision made weeks before the event?

Works cited

1. Targeting at Machine Speed: The Capabilities—and Limits—of Artificial Intelligence, https://mwi.westpoint.edu/targeting-at-machine-speed-the-capabilities-and-limits-of-artificial-intelligence/

2. Kill Chains: An Overview \- AppCheck, https://appcheck-ng.com/kill-chains-overview/

3. Joint Chiefs of Staff, Joint Publication 3-60 “Joint Targeting,” January 31, 2013\. Unclassified. \- OCR of the Document | National Security Archive, https://nsarchive.gwu.edu/media/19565/ocr

4. Weapon Target Assignment \- SciSpace, https://scispace.com/pdf/weapon-target-assignment-5779o9ikmo.pdf

5. The Kalman Filter — Foundations, Computation, and Applied Engineering — Andrey Sydelov, https://sydelov.com/de-blog/kalman-filter-foundations

6. Architecting the Information Age War \- Wavell Room, https://wavellroom.com/2021/02/23/architecting-the-information-age-war/

7. ADVANCING DECISION-CENTRIC WARFARE: \- Amazon S3, https://s3.amazonaws.com/media.hudson.org/Clark%20Patt%20Walton\_Advancing%20Decision-Centric%20Warfare.pdf

8. A Modified Bayesian Framework for Multi-Sensor Target Tracking with Out-of-Sequence-Measurements \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC7412189/

9. Sensor Fusion System Architecture: Design Patterns and Best Practices, https://sensorfusionauthority.com/sensor-fusion-architecture/

10. 2014 17th International Conference on Information Fusion (FUSION, https://www.proceedings.com/content/023/023747webtoc.pdf

11. 1097 Track-before-detect for passive radar: Detecting and tracking, https://www.researchgate.net/publication/402273521\_1097\_Track-before-detect\_for\_passive\_radar\_Detecting\_and\_tracking\_multiple\_unmanned\_aerial\_vehicles

12. Joint Angle Tracking with Inertial Sensors \- PDXScholar, https://pdxscholar.library.pdx.edu/context/open\_access\_etds/article/1660/viewcontent/ElGohary\_psu\_0180D\_10667.pdf

13. Robust Multi-Sensor Fusion for Dynamic Robotic Grasping Using Fuzzy Adaptive Extended Kalman Filter \- International Core Journal of Engineering, http://icj-e.org/index.php/ojs/article/download/319/309

14. Tutorial on Multisensor Management and Fusion Algorithms for Target Tracking \- ResearchGate, https://www.researchgate.net/profile/Mohamed\_Mourad\_Lafifi/post/Measurement-selection-of-IMM-How-to-calculate-gate-properly/attachment/59d639fd79197b80779972d7/AS%3A404182436728833%401473376010527/download/Tutorial+on+Multisensor+Management+and+Fusion+Algorithms+for+Target+Tracking.pdf

15. A modified adaptive Kalman filter algorithm for the distributed underwater multi-target passive tracking system \- AIP Publishing, https://pubs.aip.org/asa/jel/article/5/1/016001/3329102/A-modified-adaptive-Kalman-filter-algorithm-for

16. Sensor Fusion with Asynchronous Decentralized Processing for 3D Target Tracking with a Wireless Camera Network \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC9919314/

17. Multi-Sensor Data Fusion: An overview of concepts, methods and classifications \- IEEE Xplore, https://ieeexplore.ieee.org/iel8/6287639/6514899/11433650.pdf

18. A Reliability-Based Multisensor Data Fusion with Application in Target Classification \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC7218870/

19. Sensor Data Fusion for Context-Aware Computing Using Dempster-Shafer Theory, https://www.cs.cmu.edu/\~whd/publications/whd-thesis.pdf

20. Multisensor Data Fusion in IoT Environments in Dempster–Shafer Theory Setting: An Improved Evidence Distance-Based Approach \- MDPI, https://www.mdpi.com/1424-8220/23/11/5141

21. Exact and Heuristic Algorithms for the Weapon-Target Assignment Problem \- PubsOnLine, https://pubsonline.informs.org/doi/10.1287/opre.1070.0440

22. Weapon-target assignment problem: exact and approximate solution algorithms, https://ideas.repec.org/a/spr/annopr/v312y2022i2d10.1007\_s10479-022-04525-6.html

23. Real-Time Heuristic Algorithms for the Static Weapon-Target Assignment Problem \- AFIT Scholar, https://scholar.afit.edu/cgi/viewcontent.cgi?article=2646\&context=etd

24. Greedy Target Assignment with Interference Constraints Between Defensive Weapons | Journal of Aerospace Information Systems, https://arc.aiaa.org/doi/10.2514/1.I011041

25. DoD Directive 3000.09, "Autonomy in Weapon Systems," January 25, 2023 \- Executive Services Directorate, https://www.esd.whs.mil/portals/54/documents/dd/issuances/dodd/300009p.pdf

26. TRANSFORM COMMAND AND CONTROL FOR MULTIDOMAIN OPERATIONS \- Government Executive, https://events.govexec.com/the-future-of-army-fires-and-multi-domain-environments-how-data-convergence-is-galvanizing-future-weapon-systems/download/file-20230407175507-transform-command-and-control-for-multidomain-operations.pdf

27. Speed. Delivery. Agility. \- Space Development Agency, https://www.sda.mil/wp-content/uploads/2022/01/sda-about-us-2020-to-2021.pdf

28. Integrated Air & Missile Defense | Raytheon \- RTX, https://www.rtx.com/raytheon/what-we-do/integrated-air-and-missile-defense

29. LIDS Family of Systems Brochure \- SRC, Inc., https://www.srcinc.com/pdf/lids-family-of-systems-brochure.pdf

30. Not by widgets alone \- Armed Forces Journal, https://armedforcesjournal.com/not-by-widgets-alone/

31. Human-AI Teaming: State of the Art and Research Needs \- SINTEF, https://www.sintef.no/globalassets/project/hfc/documents/2021-human-ai-interaction-26355.pdf

32. A Weighted Decision-Level Fusion Architecture for Ballistic Target Classification in Midcourse Phase \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC9460598/

33. Toward a Theory of Situation Awareness in Dynamic Systems \- ResearchGate, https://www.researchgate.net/publication/345054061\_Toward\_a\_Theory\_of\_Situation\_Awareness\_in\_Dynamic\_Systems

34. Iran Air Flight 655 \- Wikipedia, https://en.wikipedia.org/wiki/Iran\_Air\_Flight\_655

35. SHOOT, DON'T SHOOT \- Air and Space Power Centre, https://airpower.airforce.gov.au/sites/default/files/2021-03/AF35-Shoot-Dont-Shoot-Minimising-the-Risk-of-Catastrophic-Error-Through-High-Consequence-Decision-Making.pdf

36. THE USS VINCENNES INCIDENT, http://real.mtak.hu/144643/1/document%20%2813%29.pdf

37. Army of None 2017053908, 9780393608984, 9780393608991 \- DOKUMEN.PUB, https://dokumen.pub/army-of-none-2017053908-9780393608984-9780393608991.html

38. EVALUATING MEANINGFUL HUMAN CONTROL WITH GENERAL SYSTEMS PERFORMANCE THEORY, https://hammer.purdue.edu/ndownloader/files/54127280

39. Army of None: Autonomous Weapons and the Future of War \- PDFDrive.com \- IS MUNI, https://is.muni.cz/el/fss/podzim2020/BSSn4411/um/Army\_of\_None\_\_Autonomous\_Weapons\_and\_the\_Future\_of\_War\_p1-245.pdf

40. Phalanx CIWS \- Wikipedia, https://en.wikipedia.org/wiki/Phalanx\_CIWS

41. R2D2 with Attitude: The Story of the Phalanx Close-In Weapons \- NavWeaps, https://www.navweaps.com/index\_tech/tech-103.php

42. USS Missouri \- laststandonzombieisland, https://laststandonzombieisland.com/tag/uss-missouri/

43. ARMY OF NONE Autonomous Weapons and the Future of War, https://ftp.idu.ac.id/wp-content/uploads/ebook/tdg/MILITARY%20PLATFORM%20DESIGN/Army%20of%20None%20Autonomous%20Weapons%20and%20the%20Future%20of%20War.pdf

44. Steam Workshop::Type 45 Destroyer \- HMS Dauntless, https://steamcommunity.com/sharedfiles/filedetails/?l=malay\&id=2638034255

45. Raytheon Lands $237 Million Counter-UAS Army Contract | Industrial Equipment News, https://www.ien.com/operations/news/22834002/raytheon-lands-237-million-counteruas-army-contract

46. Army buys 600 additional Coyote counter-drone weapons amid attacks on US troops, https://defensescoop.com/2024/02/09/army-600-coyote-counter-drone-rtx/

47. Army awards big contract for Coyote interceptors amid growing demand for counter-drone weapons | DefenseScoop, https://defensescoop.com/2024/09/27/army-awards-contract-coyote-interceptors-raytheon-counter-drone-197m/

48. Raytheon Coyote \- Wikipedia, https://en.wikipedia.org/wiki/Raytheon\_Coyote

49. Virtual System of Systems Analysis of Advanced Technologies for Joint All Domain Command and Control (JADC2) \- MODSIM World, https://www.modsimworld.org/papers/2023/MODSIM\_2023\_paper\_4853.pdf

50. Understanding vigilance and its decrement: theoretical, contextual, and neural insights, https://www.frontiersin.org/journals/cognition/articles/10.3389/fcogn.2025.1617561/full

51. Surveillance Operator Fatigue: Why Attention Fades After 30 Minutes \- EagleSight.ai, https://eaglesight.ai/resources/science-of-surveillance-fatigue

52. Chapter: 1 Introduction \- Read "Human-AI Teaming: State-of-the-Art and Research Needs" at NAP.edu, https://www.nationalacademies.org/read/26355/chapter/3

53. Scientific Measurement of Situation Awareness in Operational Testing, https://itea.org/wp-content/uploads/2023/09/Scientific-Measurement-of-Situation-Awareness-in-Operational-Testing-1.pdf

54. Situation awareness \- Wikipedia, https://en.wikipedia.org/wiki/Situation\_awareness

55. United States, Use of Autonomous Weapons \- How does law protect in war? \- ICRC, https://casebook.icrc.org/case-study/united-states-use-of-autonomous-weapons

56. DoD Announces Update to DoD Directive 3000.09, 'Autonomy In Weapon Systems', https://www.war.gov/News/Releases/Release/article/3278076/dod-announces-update-to-dod-directive-300009-autonomy-in-weapon-systems/

57. DOD Is Updating Its Decade-Old Autonomous Weapons Policy, but Confusion Remains Widespread \- CSIS, https://www.csis.org/analysis/dod-updating-its-decade-old-autonomous-weapons-policy-confusion-remains-widespread

58. DOT\&E FY2023 Annual Report \- Director Operational Test and Evaluation, https://www.dote.osd.mil/Portals/97/pub/reports/FY2023/other/2023annual-report.pdf

59. autonomous weapons, directive 3000.09, and the "appropriate levels of human judgment over the use of force \- Scholarly Publications Leiden University, https://scholarlypublications.universiteitleiden.nl/access/item%3A3761742/view

60. A model for types and levels of human interaction with automation \- ResearchGate, https://www.researchgate.net/publication/285855850\_A\_model\_for\_types\_and\_levels\_of\_human\_interaction\_with\_automation

61. A human touch: autonomous weapons, DoD Directive 3000.09 and the interpretation of 'appropriate levels of human judgment over the use of force' (Chapter 9), https://www.cambridge.org/core/books/autonomous-weapons-systems/human-touch-autonomous-weapons-dod-directive-300009-and-the-interpretation-of-appropriate-levels-of-human-judgment-over-the-use-of-force/DD3239C7CBE733DB9CC906283F18C7B8

62. By Algorithm or Order: Integrating Lethal Autonomous Weapon Systems into Targeting, https://www.armyupress.army.mil/Journals/Military-Review/Online-Exclusive/2026-OLE/Algorithm-or-Order/

63. Conscientious Centaurs \- Marine Corps University, https://www.usmcu.edu/Outreach/Marine-Corps-University-Press/MCU-Journal/JAMS-vol-16-no-2/Conscientious-Centaurs/

64. Target Tracking and Data Fusion for Ground Situational Awareness \- NATO, https://publications.sto.nato.int/publications/STO%20Educational%20Notes/STO-EN-SET-235/EN-SET-235-04.pdf