// UNCLASSIFIED // DOCTRINE LIBRARY
Defense Taxonomy.
A working reference for the protocols, frameworks, and operating concepts that define modern collaborative combat aircraft, autonomous loyal wingman teaming, electromagnetic-spectrum resilience, and distributed command-and-control software.
249 terms · 4 categories · all unclassified · public release
▎CCA Protocols
CCA & Open-Mission Protocols
Open-architecture standards and datalink protocols that govern Collaborative Combat Aircraft, modular mission systems, and crewed-uncrewed teaming.
- ANW2 Adaptive Networking Wideband WaveformANW2Commercial-derivative wideband MANET waveform widely fielded by U.S. forces.
- ARINC 653Industry standard for time- and space-partitioned avionics operating systems.
- Autonomy Reference ArchitectureARAGovernment-owned reference design for layered autonomy stacks on CCA-class platforms.
- Battlefield Airborne Communications NodeBACNHigh-altitude airborne gateway translating between disparate tactical waveforms.
- C5ISR/EW Modular Open Suite of StandardsCMOSSU.S. Army modular open standard for converged C5ISR and EW payloads on ground vehicles.
- Collaborative Combat AircraftCCAU.S. Air Force program for affordable, autonomous wingmen teamed with crewed fighters.
- Common Data LinkCDLDoD-standard high-bandwidth point-to-point ISR datalink family.
- Cursor on TargetCoTLightweight XML schema for sharing position, time, and event data across systems.
- Data Distribution ServiceDDSOMG real-time publish-subscribe middleware standard widely used in defense.
- DO-178CSoftware-considerations standard for airborne systems certification.
- Future Airborne Capability EnvironmentFACEOpen standard for portable airborne software components running on a common operating environment.
- Intra-Flight Data LinkIFDLF-22 native low-observable datalink for flight-level sensor and track sharing.
- Joint Range Extension Application ProtocolJREAPStandard for tunneling Link 16 J-series messages over IP and SATCOM.
- Joint Tactical Radio SystemJTRSDoD program family for software-defined, multi-band tactical radios.
- Link 16Jam-resistant tactical data exchange network for joint and coalition platforms.
- Loyal Wingman ConceptDoctrinal pattern of pairing autonomous uncrewed aircraft with crewed lead.
- Manned-Unmanned TeamingMUM-TDoctrinal and technical framework for cooperative crewed-uncrewed operations.
- Modular Open Systems ApproachMOSAU.S. DoD acquisition mandate requiring open architectures and standardized, replaceable modules.
- Multifunction Advanced Data LinkMADLLow-probability-of-intercept stealth datalink used by F-35 and select fifth-gen aircraft.
- Open Mission SystemsOMSU.S. Air Force government-owned interface standard for airborne mission systems.
- Sensor Open Systems ArchitectureSOSATri-service hardware-and-software open standard for sensor and EW payload modules.
- SkyborgU.S. Air Force autonomy core program that seeded the Collaborative Combat Aircraft initiative.
- Software Communications ArchitectureSCAOpen architecture for portable software-defined radio waveforms.
- Soldier Radio WaveformSRWNetworking waveform for handheld and small-form-factor tactical radios.
- STANAG 4586NATO standard interface for unmanned-aircraft control stations and platforms.
- STANAG 4660NATO interoperable command and control datalink for unmanned systems.
- Tactical Targeting Network TechnologyTTNTHigh-rate, low-latency IP-based airborne network optimized for closing kill chains.
- Team Awareness KitTAKGovernment-furnished situational awareness platform with Android, iOS, and Windows clients.
- Universal Command and Control InterfaceUCIGovernment-owned C2 messaging standard for off-board control of air and space assets.
- Vehicular Integration for C4ISR/EW InteroperabilityVICTORYU.S. Army standard for in-vehicle network and shared resources across electronic systems.
- Wideband Networking WaveformWNWHigh-throughput mobile ad-hoc networking waveform for U.S. tactical units.
▎Loyal Wingman
Autonomous Loyal Wingman Orchestration
Decision frameworks, planners, and run-time assurance patterns used to orchestrate autonomous wingmen and swarming agents under commander's intent.
- A* PathfindingBest-first graph search algorithm with admissible heuristics.
- ACAS XuAirborne Collision Avoidance System variant designed for unmanned aircraft.
- Auction-Based Task AllocationMarket-style mechanism for assigning tasks to agents by bid.
- Behavior TreesHierarchical control structure for composing autonomous agent behaviors.
- Byzantine Fault ToleranceBFTProperty of agreeing despite arbitrary, including malicious, node behavior.
- Commander's Intent EncodingMachine-readable representation of mission purpose and acceptable actions.
- Consensus AlgorithmsDistributed protocols ensuring agreement among nodes despite faults.
- Contract Net ProtocolFoundational task-announcement, bid, and award protocol for multi-agent systems.
- Cooperative Engagement CapabilityCECU.S. Navy system for fusing sensor tracks across ships and aircraft into a single composite track.
- Digital Twin (Autonomy)High-fidelity simulation paired with the live system for rehearsal and assurance.
- Finite State MachinesFSMClassical state-and-transition control model for bounded autonomous behavior.
- Goal-Oriented Action PlanningGOAPLightweight STRIPS-style planner that chains actions by precondition and effect.
- Hierarchical Task NetworksHTNPlanning paradigm that decomposes tasks into ordered subtasks via methods.
- Joint Probabilistic Data AssociationJPDATracking technique that probabilistically associates measurements with tracks.
- Kalman FilterOptimal recursive estimator for linear-Gaussian dynamic systems.
- Leader ElectionDistributed primitive for selecting a single coordinator among peers.
- Multiple Hypothesis TrackingMHTTracking algorithm that maintains parallel hypotheses about data association.
- Particle FilterSequential Monte Carlo estimator for nonlinear, non-Gaussian state estimation.
- Rapidly-Exploring Random TreeRRTSampling-based motion planner for high-dimensional configuration spaces.
- Run-Time AssuranceRTASafety architecture that monitors and overrides untrusted autonomy at run time.
- Self-Healing MeshProperty of a mesh network to reroute around lost links and nodes automatically.
- Sense-and-AvoidSAA / DAACapability for uncrewed aircraft to detect and avoid other airspace users.
- Sim-to-Real TransferSet of techniques for porting policies trained in simulation to physical hardware.
- Simplex ArchitectureRun-time assurance pattern with a verified backup controller and decision monitor.
- Swarm Cohesion (Flocking)Local rule sets producing emergent group motion in multi-agent fleets.
- Utility AIDecision pattern that scores candidate actions by weighted utility curves.
▎EW Mesh
EW-Resilient Mesh Network Topologies
Mobile ad-hoc networking, cognitive radio, and electromagnetic-spectrum tradecraft used to keep tactical meshes alive in contested environments.
- Ad-Hoc On-Demand Distance VectorAODVReactive routing protocol that builds routes on demand.
- Alternative Positioning, Navigation, and TimingAlt-PNTNon-GNSS sources of PNT used when satellite navigation is denied or degraded.
- Anti-Jam TechniquesAJWaveform and protocol techniques that preserve link integrity under deliberate jamming.
- BeamformingCoherent steering of antenna-array energy to a specific spatial direction.
- Cognitive RadioCRRadio that senses its RF environment and adapts waveform, frequency, and power autonomously.
- Controlled Reception Pattern AntennaCRPAAdaptive antenna array that nulls jammers by steering its reception pattern.
- Destination-Sequenced Distance-VectorDSDVProactive distance-vector protocol with sequence numbers to prevent routing loops.
- Direct-Sequence Spread SpectrumDSSSSpread-spectrum technique that multiplies the data signal by a high-rate pseudo-noise code.
- Dynamic Spectrum AccessDSAReal-time, opportunistic use of spectrum based on sensing and policy.
- Electromagnetic Battle ManagementEMBMReal-time C2 of friendly EM emissions and adversary spectrum activity.
- Electromagnetic Spectrum OperationsEMSOJoint doctrine integrating electronic warfare and spectrum management as a single discipline.
- Electronic AttackEAUse of EM energy to degrade, neutralize, or destroy adversary capability.
- Electronic ProtectionEPMeasures that protect friendly use of the EM spectrum from EA.
- Frequency-Hopping Spread SpectrumFHSSSpread-spectrum technique that rapidly changes carrier frequency over a wide band.
- GNSS Anti-JamHardware and software techniques that preserve GNSS positioning under jamming.
- Have QuickU.S. UHF anti-jam frequency-hopping waveform for air-to-air and air-to-ground voice.
- Inertial Navigation SystemINSSelf-contained PNT computed from accelerometers and gyroscopes.
- Joint Tactical Radio SystemJTRSU.S. DoD program that established the software-defined radio and waveform portability framework.
- Link 16NATO tactical datalink for situational awareness and weapons coordination.
- Low Probability of InterceptLPIWaveform property that minimizes the chance of detection by adversary signals intelligence.
- Mobile Ad-Hoc NetworkMANETSelf-forming, self-healing IP network where every node is also a router.
- Multifunction Advanced Data LinkMADLStealth-compatible LPI/LPD datalink developed for fifth-generation aircraft.
- Multiple-Input Multiple-OutputMIMOAntenna technique using multiple transmit and receive elements for capacity and resilience gains.
- Optimized Link State RoutingOLSRv2Proactive link-state routing protocol designed for MANETs.
- Single Channel Ground and Airborne Radio SystemSINCGARSLegacy U.S. VHF combat-net radio family with FHSS anti-jam mode.
- Software-Defined RadioSDRRadio whose physical-layer behavior is defined in software, enabling waveform reprogramming.
- Soldier Radio WaveformSRWU.S. government-owned narrowband mesh waveform for dismounted forces.
- Spectrum PolicyMachine-readable rules governing what a radio may transmit, when, and where.
- Tactical Scalable Mobile Ad-Hoc NetworkTSMPersistent Systems waveform optimized for high-density, mobile mesh at the tactical edge.
- Tactical Targeting Network TechnologyTTNTHigh-throughput, low-latency airborne IP datalink for time-critical targeting.
- Wideband Networking WaveformWNWJTRS high-throughput mesh waveform for vehicle and command-post backbone.
▎Distributed C2
Distributed C2 Software Frameworks
Software frameworks, mission systems, and joint-all-domain command architectures that distribute decision authority across the kill web.
- Advanced Battle Management SystemABMSU.S. Air Force JADC2 implementation centered on data fabric and edge compute.
- AI Fires RecommenderSoftware service that pairs detected targets with available shooters and effects.
- Commander's IntentConcise statement of purpose, key tasks, and end state that guides subordinate initiative.
- Common Operating PictureCOPSingle, shared, real-time view of the battlespace across echelons and partners.
- Common Tactical PictureCTPEchelon-specific subset of the COP focused on the tactical fight.
- Continuous Authority to OperatecATOAccreditation model granting ongoing ATO based on continuous monitoring and pipeline controls.
- Cooperative Engagement CapabilityCECNaval sensor-netting system that fuses radar tracks across ships into a single fire-control picture.
- Cursor-on-TargetCoTLightweight XML message standard for exchanging tactical events.
- Data Distribution ServiceDDSOMG publish-subscribe middleware standard widely used in defense systems.
- DevSecOpsPractice of integrating security into continuous software delivery pipelines.
- Distributed Decision-MakingOperational pattern in which decision authority is partitioned across echelons and nodes.
- Distributed Maritime OperationsDMONavy operating concept dispersing combat power across many networked platforms.
- Human-in-the-LoopControl pattern requiring explicit human authorization for each consequential action.
- Human-on-the-LoopSupervisory control pattern where the human can intervene but does not act on every step.
- Integrated Tactical NetworkITNArmy initiative blending purpose-built and commercial networking at the tactical edge.
- Joint All-Domain Command and ControlJADC2U.S. DoD warfighting concept connecting sensors and shooters across every domain.
- Joint Battle Command-PlatformJBC-PU.S. Army mounted/dismounted C2 platform replacing FBCB2/Blue Force Tracker.
- Kubernetes at the EdgeUse of container orchestration to deploy and manage C2 services on tactical hardware.
- Mission CommandCommand philosophy granting subordinates intent-based discretion under decentralized authority.
- Platform OneU.S. Air Force enterprise DevSecOps platform serving the joint force.
- Project ConvergenceU.S. Army JADC2 campaign of learning that demonstrates sensor-to-shooter timelines.
- Project OvermatchU.S. Navy JADC2 program focused on networks, AI, and distributed maritime operations.
- Real-Time Publish-Subscribe ProtocolRTPSWire protocol underlying DDS, providing interoperable pub-sub over IP.
- Rules of EngagementROEDirectives that define the circumstances under which force may be used.
- Sensor FusionCombining multi-sensor data into a unified, higher-confidence track picture.
- Tactical Assault KitTAKGovernment-owned situational awareness and C2 software family across services.
- Tactical Data FabricDistributed data plane providing discovery, transport, and policy across the kill web.
- Zero Trust ArchitectureZTASecurity model that authenticates and authorizes every request regardless of network location.
▎AI & Multi-Agent
AI-Powered Defense Technology & Multi-Agent Systems
Foundation models, agentic architectures, multi-agent learning, and trusted-execution patterns that let KhanBMS distribute intelligence across the edge.
- Acoustic Signature ClassificationAI classification of vehicles, aircraft, weapons, or activity from sound and vibration patterns.
- Adversarial Machine LearningAMLStudy and defense of attacks that manipulate AI through crafted inputs, poisoned data, or model theft.
- Adversarial PromptingInputs designed to coerce a language model or agent into unsafe, unauthorized, or false behavior.
- Agent MemoryPersistent short-term and long-term context stores used by agents to remember facts, goals, actions, and lessons.
- Agent-to-Agent ProtocolA2ACommunication pattern for autonomous agents to negotiate tasks, exchange state, and request support.
- Agentic AIAI systems that pursue goals through planning, tool use, memory, and feedback loops rather than single-shot inference.
- AI Bill of MaterialsAIBOMInventory of models, datasets, adapters, tools, dependencies, licenses, and provenance in an AI system.
- AI Change DetectionDetection of meaningful differences across images, maps, sensor passes, or operational data over time.
- AI Data FabricIntegrated data layer that connects operational, sensor, model, metadata, and governance sources for AI workflows.
- AI Object TrackingMachine-learning methods that maintain object identity and trajectory across frames, sensors, and time.
- AI Red TeamingStructured adversarial testing of AI systems to expose unsafe, biased, exploitable, or brittle behavior.
- AI Risk Management FrameworkAI RMFStructured approach to identifying, measuring, managing, and governing AI risks across the lifecycle.
- AI Sensor FusionMachine-learning methods that combine multiple sensor streams into a better estimate than any source alone.
- AI Sensor TaskingAI selection of where, when, and how sensors should collect next to reduce uncertainty.
- AI Supply Chain SecurityProtection of datasets, weights, code, dependencies, tooling, and deployment pipelines for AI systems.
- AI Trusted Execution EnvironmentAI-TEEHardware-isolated environment for protecting model weights, inputs, and inference outputs from a compromised host.
- AI WargamingUse of AI agents and simulations to explore adversary moves, blue responses, and campaign dynamics.
- AI WatermarkingEmbedding or detecting signals that identify AI-generated content or model ownership.
- Automatic Target RecognitionATRAI-enabled detection and classification of objects, vehicles, emitters, or activities from sensor data.
- Autonomous CEMACEMA-AIAI coordination of cyber and electromagnetic activities as an integrated operational effect.
- Autonomous Cyber DefenseAI systems that detect, triage, contain, and respond to cyber threats with bounded automation.
- Autonomous Electronic WarfareAEWAI-assisted sensing, decision, and waveform control for electronic attack, protection, and support.
- Autonomous LogisticsAI and robotic coordination of supply, maintenance, routing, and distribution with minimal manual control.
- Autonomous Undersea SystemsAUSAI-enabled underwater vehicles and sensors operating with sparse communications and harsh navigation constraints.
- Autonomy Test and EvaluationT&ETest discipline for validating autonomous systems across simulation, hardware, field trials, and adversarial scenarios.
- Belief-Desire-Intention AgentsBDIAgent model that separates what an agent believes, wants, and intends to do.
- Blackboard ArchitectureAI system pattern where independent specialists read and write to a shared problem workspace.
- Centralized Training, Decentralized ExecutionCTDETraining pattern where agents learn with shared global information but deploy using local observations.
- Cislunar AutonomyAutonomous navigation, coordination, and operations for spacecraft and infrastructure between Earth and the Moon.
- Confidence CalibrationEnsuring model confidence scores correspond to real-world likelihood of being correct.
- Confidential AI ComputingUse of encryption, enclaves, and attestation to protect AI workloads while data is in use.
- Consensus Algorithms for AIProtocols that let distributed AI nodes agree on shared state, leaders, or decisions despite latency and loss.
- Constitutional AICAIAlignment approach where model behavior is shaped by written principles and self-critique instead of only human labels.
- Cooperative PerceptionShared perception where multiple platforms combine local observations to improve detection and tracking.
- Counter-AI OperationsActions that detect, disrupt, deceive, or exploit adversary AI systems and data pipelines.
- Counter-UAS AIC-UAS AIAI methods for detecting, classifying, tracking, prioritizing, and defeating uncrewed aerial threats.
- Course-of-Action GenerationCOA AIAI generation and comparison of plausible mission options under constraints and commander intent.
- Data PoisoningAttack that corrupts training or fine-tuning data to implant bad behavior or degrade performance.
- Decision Support AIDSAIAI systems that synthesize data, alternatives, risk, and explanations for commanders or operators.
- Deepfake DetectionAI identification of synthetic or manipulated audio, video, imagery, and documents.
- Defense Foundation ModelsDFMLarge pretrained AI models adapted for military planning, perception, language, and decision-support workloads.
- Digital Twin SimulationLive or synchronized synthetic replica of a platform, unit, network, or environment used for testing and rehearsal.
- Distributed AuctionDecentralized bidding process used to assign tasks across agents without a single allocation server.
- Doctrine-Grounded ReasoningAI reasoning grounded in authoritative doctrine, tactics, ROE, and unit-specific operating procedures.
- Domain RandomizationTraining technique that varies simulated conditions widely so models generalize better to reality.
- Edge InferenceRunning AI models on tactical hardware at the point of sensing or action instead of relying on distant cloud compute.
- Edge Model RegistryVersioned catalog that tracks which models, adapters, signatures, and policies are deployed to tactical nodes.
- Effects-Based Planning AIAI planning that reasons from desired operational effects back to actions, assets, and timing.
- Emergent CommunicationLearned communication protocols that arise between agents during training rather than being hand-designed.
- EO/IR FusionEO/IRFusion of visible and infrared imagery to improve recognition across day, night, weather, and obscurants.
- Evasion AttacksInputs crafted at inference time to make a model misclassify or choose the wrong action.
- Event-Based VisionNeuromorphic camera processing that reacts to pixel-level brightness changes instead of full frames.
- Explainable AIXAIMethods that show why an AI system produced a prediction, recommendation, or action.
- Fault-Tolerant InferenceAI inference designed to keep functioning despite node loss, degraded sensors, hardware faults, or link outages.
- Federated LearningFLTraining approach where nodes learn from local data and share updates instead of raw data.
- Fires Recommendation AIAI decision aid that recommends potential fire missions, timing, assets, and collateral-risk checks.
- Flocking AlgorithmsRules that generate coherent group motion through separation, alignment, and cohesion behaviors.
- Formation ControlAlgorithms that maintain relative positions, spacing, and geometry across autonomous vehicles.
- FPGA ML AccelerationFPGA-AIUse of field-programmable gate arrays to run low-latency or reconfigurable AI workloads.
- Graph-of-Thought ReasoningGoTReasoning method that models intermediate ideas as a graph so steps can merge, revise, or cross-check each other.
- Ground Robotics AutonomyGRAAI control and perception for unmanned ground vehicles, robotic mules, breachers, and urban scouts.
- Hierarchical Task Networks for AIHTN-AIPlanning formalism that decomposes abstract missions into ordered executable tasks and methods.
- Human-on-the-Loop AI SupervisionHOTL-AISupervisory control model where humans monitor autonomous systems and can intervene or abort.
- Hyperspectral Target DetectionHSIAI analysis of many spectral bands to identify materials, camouflage, or disturbed terrain.
- Intent InferenceAI estimation of friendly, neutral, or adversary intent from behavior, context, and prior patterns.
- Interpretable Machine LearningIMLModeling approaches whose internal logic can be understood directly or with minimal post-hoc explanation.
- Jailbreak ResistanceDefenses that stop users or inputs from bypassing model safety and policy constraints.
- Knowledge Graph ReasoningKGRReasoning over entities, relationships, provenance, and constraints represented as a graph.
- Large Language Models for DefenseLLMTransformer language models tuned for doctrine search, staff workflows, planning assistance, and machine-readable orders.
- League-Based TrainingSelf-play method that maintains a population of opponents and teammates to improve robustness.
- LiDAR PerceptionAI interpretation of point clouds for detection, tracking, terrain mapping, and navigation.
- LLM Orchestration LayerMiddleware that routes models, prompts, tools, memory, retrieval, policy, and telemetry across AI workflows.
- Logistics Optimization AIAI optimization of supplies, spares, fuel, batteries, routes, and maintenance across distributed forces.
- Low-Rank AdaptationLoRAFine-tuning technique that updates small rank-decomposition matrices instead of all model weights.
- Loyal Wingman AIAutonomy stack that lets an uncrewed aircraft cooperate with crewed aircraft under mission intent.
- Market-Based Task Allocation for AITasking method where agents bid for work based on cost, capability, risk, and availability.
- Mean-Field Reinforcement LearningMFRLApproximation method for learning in very large agent populations by modeling aggregate neighbor behavior.
- Mechanistic InterpretabilityAnalysis of internal neural-network circuits, features, and representations to understand model behavior.
- Mission Data LakehouseUnified storage architecture for raw, structured, and analytic mission data used by AI and operators.
- Mission Planning AgentAI agent that helps generate, revise, and monitor mission plans under commander intent and constraints.
- Mixture of ExpertsMoEModel architecture that activates specialized subnetworks for different tokens or tasks to scale capability efficiently.
- MLOps for DefenseMLOps-DLifecycle practices for building, testing, approving, deploying, monitoring, and updating military AI.
- Model Cards for DefenseDocumentation artifacts describing model purpose, training data, metrics, limits, and approved uses.
- Model Context ProtocolMCPOpen protocol pattern for exposing tools, resources, and prompts to model agents through standard interfaces.
- Model DistillationKDTraining method that transfers behavior from a larger teacher model into a smaller deployable student model.
- Model ExtractionAttack that recreates or approximates a model by querying it and observing outputs.
- Model InversionAttack that infers sensitive training data or attributes from model outputs or gradients.
- Model ObservabilityMonitoring of model inputs, outputs, drift, latency, confidence, and failures after deployment.
- Model PartitioningDividing model layers or experts across devices so inference can run over a distributed system.
- Model QuantizationINT8/INT4Reducing model numerical precision to cut memory, latency, and power while preserving enough accuracy.
- Multi-Agent DebateMADTechnique where multiple model agents argue, critique, and revise answers before a decision is surfaced.
- Multi-Agent Reinforcement LearningMARLReinforcement-learning framework where multiple agents learn cooperative or adversarial behavior together.
- Multimodal Foundation ModelsMFMFoundation models that jointly process text, imagery, video, audio, maps, and structured sensor data.
- Multimodal Sensor FusionFusion of data across different sensing modalities, including imagery, RF, acoustic, cyber, text, and tracks.
- Neural Processing Unit AcceleratorsNPUSpecialized chips for accelerating neural-network inference on edge and embedded devices.
- Neural Radiance FieldsNeRFNeural scene representation that reconstructs 3D views from multiple images or sensor positions.
- Neuromorphic ComputingBrain-inspired hardware and algorithms that process spikes or events for low-power AI.
- On-Device Fine-TuningLocal adaptation of AI models on tactical devices using recent mission or environment data.
- OODA Loop AccelerationUse of AI to compress observe-orient-decide-act cycles while preserving human judgment and control.
- Operational Anomaly DetectionAI detection of unusual platform behavior, network activity, sensor patterns, or adversary activity.
- Parameter-Efficient Fine-TuningPEFTFamily of methods that customize large models by training a small fraction of parameters.
- Plan-and-Execute AgentsAgent pattern that separates high-level planning from stepwise execution and monitoring.
- Policy GuardrailsDeterministic and model-assisted controls that constrain what AI systems may say, decide, or execute.
- Predictive Maintenance AIMachine learning that forecasts equipment failure and maintenance needs from telemetry and history.
- Prompt Engineering for DefenseDesign of instructions, context, and constraints that steer model behavior for military workflows.
- Prompt Injection DefenseControls that prevent untrusted text or content from overriding a model agent’s system instructions or tools.
- ReAct Agent PatternReActReasoning-and-acting pattern where an AI alternates between thought, tool call, observation, and next action.
- Reinforcement Learning from Human FeedbackRLHFAlignment method that uses human preference data to shape model behavior after pretraining.
- Responsible AI for DefenseRAIGovernance practices that align military AI with lawful, ethical, reliable, and accountable use.
- RF FingerprintingMachine-learning identification of devices or emitters from subtle radio-frequency signal characteristics.
- Risk-Aware PlanningPlanning that explicitly models uncertainty, loss, detection, collateral risk, and mission failure probabilities.
- ROE-Aware PlanningROE AIPlanning methods that encode rules of engagement, authorities, and escalation constraints.
- Role AssignmentAlgorithmic allocation of scout, relay, decoy, strike, and reserve roles across autonomous assets.
- Run-Time Assurance for AIRTA-AISafety architecture that monitors AI outputs and switches to a verified fallback when behavior leaves bounds.
- Secure Model ProvenanceCryptographic and procedural evidence tracking where a model, adapter, dataset, or artifact came from.
- Self-Play TrainingTraining method where agents improve by competing or cooperating against versions of themselves.
- Semantic SLAMSLAMSimultaneous localization and mapping enriched with object labels, terrain classes, and mission-relevant semantics.
- Simulation-to-Real AISim2RealTechniques that transfer AI behavior trained in simulation into physical platforms and real operations.
- Small Language ModelsSLMCompact language models optimized for local inference on constrained tactical hardware.
- Sovereign AI ModelsSAIModels trained, hosted, and governed under national or coalition control rather than foreign commercial dependency.
- Split InferenceInference architecture that divides a model between edge devices and more capable local or remote compute.
- StigmergyIndirect coordination where agents communicate by modifying or reading traces in the environment or shared state.
- Swarm IntelligenceCollective behavior emerging from many local agents rather than a single central controller.
- Synthetic Aperture Radar AISAR-AIMachine learning for interpreting SAR imagery, including detection, segmentation, and change analysis.
- Synthetic Pretraining DataSPDMachine-generated or simulated data used to expand training corpora where real examples are scarce or sensitive.
- Synthetic Training EnvironmentsSTEGenerated or simulated worlds used to train AI policies, perception models, and human teams.
- Tactical AI ComputeRuggedized compute stack for running AI on vehicles, aircraft, radios, command posts, and soldier systems.
- Target Prioritization AIAI ranking of targets by threat, value, vulnerability, timing, and mission relevance.
- TinyMLMachine learning designed for microcontrollers and ultra-low-power embedded devices.
- Tool-Use AgentsAgents that call external APIs, databases, simulators, sensors, or effectors to accomplish tasks.
- Tree-of-Thought ReasoningToTPrompting and search method that explores multiple reasoning branches before choosing an answer or plan.
- Vector DatabaseDatabase optimized for storing embeddings and retrieving semantically similar text, images, or events.
- Vision-Language ModelsVLMMultimodal models that jointly interpret imagery and language for visual question answering and scene explanation.