Abstract: Modern multi-domain operations demand kill chain closure times measured in seconds rather than minutes. This paper presents the architecture and simulation methodology of c4isr-killchain-sim, a high-fidelity computational framework that models the complete C4ISR kill chain from sensor detection through battle damage assessment. The simulation spans seven interdependent phases—sensor network fusion, C2 communications, track management, command and control decision engines, kill chain timing, JADC2 integration, and scenario execution—each implemented with physics-based latency models, resource allocation algorithms, and doctrine-aware decision logic. We characterize sensor-to-shooter latency distributions across operational scenarios, quantify the impact of communications degradation on kill chain closure probability, and demonstrate how Joint All-Domain Command and Control (JADC2) architectures transform linear kill chains into resilient kill webs. Validation against published engagement timeline benchmarks shows the model reproduces per-phase latency budgets within 8–15% of operational data across Pacific, European, and Middle Eastern threat environments.
The kill chain—the sequence of steps from detecting a threat to assessing the effect of a response—is the fundamental temporal unit of military engagement. As threat velocities increase and operating domains multiply, the time available for kill chain closure shrinks, while the number of potential sensor-shooter pathways expands. The U.S. Department of Defense's Joint All-Domain Command and Control (JADC2) concept recognizes that traditional linear kill chains, optimized for single-domain engagements, are insufficient for the speed and complexity of modern multi-domain operations [1, 2].
This paper describes the architecture and simulation methodology of c4isr-killchain-sim, a high-fidelity computational framework that models the complete C4ISR (Command, Control, Communications, Computers, Intelligence, Surveillance, and Reconnaissance) kill chain. The simulation decomposes the kill chain into seven interdependent phases, each implemented with physics-based models, resource-constrained algorithms, and doctrine-aware decision logic. By composing these phases into an end-to-end simulation, the framework enables analysis of sensor-to-shooter latency, kill chain closure probability under contested conditions, and the marginal benefit of JADC2 architectural investments.
The remainder of this paper is organized as follows. Section 2 reviews the F2T2EA kill chain framework. Sections 3–6 detail the four core simulation phases: sensor fusion, communications, track management, and C2 decision engines. Section 7 presents kill chain timing analysis. Section 8 addresses JADC2 integration and the kill web concept. Section 9 describes the operational scenarios. Section 10 discusses validation, and Section 11 concludes.
The Find-Fix-Track-Target-Engage-Assess (F2T2EA) framework, codified in Joint Publication 3-01 [3], defines six sequential phases for any engagement:
| Phase | Function | Key Metrics |
|---|---|---|
| Find | Detect presence of a threat entity | Probability of detection ($P_d$); detection range |
| Fix | Resolve location and kinematics | Position accuracy (CEP); track initiation time |
| Track | Maintain continuous state estimate | Track continuity; covariance growth rate |
| Target | Select and prioritize for engagement | Classification confidence; prioritization latency |
| Engage | Commit weapon to intercept | Weapon-target pairing; flight time |
| Assess | Evaluate engagement outcome | BDA probability; re-engagement decision time |
In a traditional linear kill chain, each phase must complete before the next begins. The total kill chain time is therefore the sum of individual phase latencies:
In practice, phases overlap and iterate, particularly the Target–Engage–Assess cycle under shoot-look-shoot doctrine (Section 6.3).
Kill chain closure probability is the likelihood that all six phases complete before the threat achieves its objective. For a threat with time-to-objective $T_{\text{obj}}$, the closure probability is:
Where individual phase latencies $t_i$ are stochastic, drawn from phase-specific distributions that depend on sensor performance, network conditions, and decision timelines. The simulation models each ti as a random variable with parameters determined by the operational context, enabling Monte Carlo estimation of $P_{\text{close}}$ across thousands of engagement instances.
A critical insight is that $P_{\text{close}}$ is not merely additive in per-phase success probabilities. Phase dependencies—for example, track quality affecting targeting confidence—create coupling that must be modeled explicitly. The simulation captures these dependencies through a phase-graph execution model where each phase receives inputs from its predecessors and produces outputs that condition downstream phases.
The sensor_net phase models four classes of sensor platforms, each contributing distinct detection and tracking capabilities:
| Sensor Class | Representative System | Primary Band | Key Capability |
|---|---|---|---|
| Space-based IR | SBIRS | Infrared | Early warning; boost-phase detection |
| Ground-based radar | EWR (UEWR) | UHF/VHF | Long-range detection; early track |
| Ship-based radar | Aegis SPY-1/6 | S-band | Precision track; fire control quality |
| Air-based | E-2D / F-35 | Multiple | Over-the-horizon; low-observable track |
Each sensor platform is parameterized by its detection probability as a function of range, aspect angle, and target radar cross-section (RCS). The simulation implements the standard radar range equation for active sensors:
and IR detection models for passive sensors incorporating atmospheric transmission, clutter, and sensor noise-equivalent temperature difference (NETD).
Coverage maps are computed as geodesic grids where each cell accumulates the composite detection probability across all contributing sensors. Cells with $P_d$ below a configurable threshold are flagged as coverage gaps, enabling sensor tasking recommendations (Section 3.3).
Multi-sensor fusion requires correlating detections from disparate platforms into unified tracks. The simulation implements a two-stage process:
Track-to-track correlation. Using a gating function based on Mahalanobis distance, candidate track pairs are tested for statistical association:
where x̂i, x̂j are state estimates from sensors i and j, Pi, Pj are their associated covariances, and the threshold is drawn from the χ² distribution at significance level α with nx degrees of freedom.
Covariance intersection (CI). When cross-covariance between correlated tracks is unknown—the typical case for multi-platform data—the simulation employs the Covariance Intersection algorithm [4] rather than naive Kalman fusion, which would underestimate uncertainty:
where $\omega \in [0, 1]$ is selected to minimize the determinant (or trace) of $P_{CI}$, guaranteeing consistency regardless of the unknown cross-correlation. This is essential for fusion across classification boundaries where shared process noise makes independence assumptions invalid.
Sensor resources are finite. The simulation models tasking as a constrained optimization problem: maximize the sum of track quality improvements across all active tracks, subject to sensor field-of-view, duty cycle, and energy constraints. The objective function for sensor s assigned to track set Ts is:
where wt is a priority weight based on threat level, ΔQs,t is the expected quality improvement from sensor s observing track t, rs,t is the resource cost, and Rs is the total resource budget for sensor s. This is cast as a binary integer program and solved approximately using a greedy heuristic with local search, trading optimality for the sub-second solution times required by the simulation loop.
The comms phase models the transport of sensor data and C2 messaging across the space data network. Optical Communication Terminal (OCT) laser links provide high-bandwidth inter-satellite connectivity with data rates of 1–10 Gbps and latency dominated by propagation and on-board processing:
where d is the slant range, c is the speed of light, and tproc includes encoding, decoding, and pointing acquisition. Mesh routing is modeled using a shortest-path algorithm over the satellite constellation graph, with link weights reflecting current queue depth and atmospheric availability for ground-space links.
Three primary RF data link families are modeled:
| Link | Band | Range | Data Rate | Application |
|---|---|---|---|---|
| Link 16 | L-band (969–1206 MHz) | ~300 nm LOS | Up to 115 kbps | Tactical data; situational awareness |
| JREAP | Various (IP-based) | Beyond LOS (satellite) | Variable | Extended range; track forwarding |
| MADL | Ku-band | ~250 nm LOS | Up to 1 Mbps | Low-observable; anti-jam |
Each link type is modeled with a per-hop latency distribution, packet loss probability, and anti-jam margin. The simulation captures the heterogeneous nature of operational networks where data traverses multiple link types between source and consumer.
End-to-end latency for a message traversing the network is the sum of propagation, transmission, processing, and queuing delays:
Priority queuing is modeled using an M/G/1 queue with four precedence levels (FLASH, IMMEDIATE, PRIORITY, ROUTINE), approximating MIL-STD-188-220 precedence handling. Higher-precedence messages receive non-preemptive priority, reducing their expected queuing delay at the expense of lower-precedence traffic. The model captures the operational reality that C2 fire-control messages experience significantly lower queuing delays than routine status updates, even over the same physical links.
Communications under electronic attack experience degradation modeled as a function of jammer power, distance, and link margin. The effective signal-to-noise ratio under jamming is:
where Jnoise is determined by the jammer's effective radiated power, the jamming path loss, and the link's processing gain (spread-spectrum factor). When SNReff falls below the minimum threshold for the link's modulation and coding scheme, the link is declared degraded and traffic is rerouted through the mesh. The simulation models progressive degradation: links first reduce their effective data rate (falling back to more robust but slower modulation), then experience increased packet loss, and finally become unavailable.
Network resilience is assessed by measuring the fraction of high-priority messages that arrive within their latency budget under increasing levels of jamming, producing a resilience curve that quantifies the operational benefit of redundant paths and multi-link architectures.
The track phase implements a multi-domain track management architecture consistent with C2BMC and Aegis combat system practices. Each detected object is assigned a Global Identifier (GID) that persists across sensor handoffs, domain boundaries, and track number changes. The composite track merges contributions from all sensors reporting on that GID, with a track quality metric $Q \in [0, 1]$ that aggregates position accuracy, velocity estimation confidence, and classification certainty:
where $\alpha, \beta, \gamma$ are doctrinal weights (default: 0.4, 0.3, 0.3), $\text{CEP}$ is the circular error probable of the position estimate, $\sigma_v$ is velocity uncertainty, and $P_{\text{class}}$ is the classification confidence. Tracks must exceed a minimum quality threshold (configurable, default $Q \ge 0.6$) to be eligible for engagement targeting.
Track handoff between domains—for example, from a space-based SBIRS track to a ground-based EWR, then to an Aegis ship—is a critical source of latency and track corruption. The simulation models handoff as a three-step process:
Handoff latency includes the communication delay for transmitting the track data, the sensor reacquisition time (which depends on the uncertainty ellipse volume relative to the sensor's field of view), and the correlation processing time. The simulation tracks cumulative handoff count and latency per track, as excessive handoffs can degrade track quality beyond the engagement threshold.
Not all sensors provide range. The simulation includes a bearing-only Target Motion Analysis (TMA) module for passive sensors that measure azimuth and elevation but not range. TMA is modeled as an extended Kalman filter with a modified polar coordinate state representation [5], which provides better observability properties than Cartesian representations for bearing-only scenarios. Convergence time depends on sensor-target geometry, with observability improving when the sensor platform maneuvers to create bearing rate.
Midcourse discrimination—distinguishing re-entry vehicles (RVs) from decoys and penetration aids—is modeled as a classification sub-problem within the track phase. Each object receives a discrimination score based on IR signature (SBIRS), radar cross-section variation, and kinematic consistency. The simulation implements a sequential likelihood ratio test:
where yn is the n-th observation vector. Classification is declared when Λn crosses an upper threshold (RV) or lower threshold (decoy), or remains in the indecisive region at the decision deadline, in which case the object is classified based on the current ratio and a conservative doctrinal bias toward treating unknowns as threats.
The c2 phase implements a Rules of Engagement (ROE) engine that gates every engagement decision. The ROE state machine operates in three modes:
| ROE State | Definition | Engagement Permission |
|---|---|---|
| WEAPONS FREE | Engage any hostile-identified track | Automatic upon hostile classification |
| WEAPONS TIGHT | Engage only tracks meeting specific criteria | Requires track classification match and positive ID |
| WEAPONS HOLD | Do not engage except in self-defense | Self-defense only; requires explicit authorization |
Transitions between ROE states are triggered by doctrinal events (threat escalation, commander directive, battle damage assessment) and are modeled with stochastic timing to capture human decision latency. The simulation allows scenario-specific ROE configurations and tracks the fraction of valid engagement opportunities that were not acted upon due to ROE constraints—a measure of the "ROE cost" in terms of missed engagements.
Weapon-Target Assignment (WTA) is formulated as a combinatorial optimization problem: assign weapons to targets to maximize expected kills, subject to weapon availability, engagement geometry, and ROE constraints. The objective is:
where $v_j$ is the value of target $j$, $W_{i,j}$ is the set of weapons assigned to target $j$ from shooter $i$, and $P_{k,j}$ is the single-shot kill probability of weapon k against target j. WTA is NP-hard [6], and the simulation implements three solution approaches:
The simulation supports multiple engagement doctrines, each with distinct temporal and resource implications:
The simulation models decision timelines for each C2 action (classification, ROE check, WTA computation, fire command) with stochastic durations drawn from log-normal distributions calibrated to operational data. Mean decision times range from 0.5 seconds for automated fire-control decisions to 30+ seconds for ROE escalation requests requiring human authorization.
Battle rhythm is modeled as the cyclic planning process that governs target nomination and weapon allocation. The Air Tasking Order (ATO) cycle (typically 72 hours) sets the strategic allocation, while dynamic targeting processes handle time-sensitive targets within the kill chain timeline. The simulation allows targets to transition between planned and dynamic targeting based on their time-sensitivity classification.
The killchain phase instruments each F2T2EA step with timing, producing a per-phase latency breakdown that can be analyzed across Monte Carlo runs. Table 1 shows representative latency distributions for a Pacific theater scenario with no communications degradation:
| Phase | Mean (s) | Std Dev (s) | P95 (s) | Dominant Delay Source |
|---|---|---|---|---|
| Find | 2.1 | 1.8 | 5.4 | Sensor scan period |
| Fix | 1.4 | 0.9 | 3.1 | Detection-to-track initiation |
| Track | 0.6 | 0.3 | 1.2 | Track quality threshold crossing |
| Target | 3.2 | 2.1 | 7.8 | Classification + WTA computation |
| Engage | 1.8 | 0.7 | 3.0 | Fire control + launch |
| Assess | 5.5 | 3.4 | 12.1 | BDA sensor tasking + assessment |
| Total | 14.6 | 5.2 | 24.8 | — |
The Assess phase is the single largest latency contributor, driven by the time required to re-task a sensor (often a different platform than the one providing fire-control quality track) and render a BDA judgment. Under shoot-look-shoot doctrine, this phase directly determines whether a second engagement is possible within the threat's time-to-objective.
Closure probability is estimated via Monte Carlo sampling of the per-phase latency distributions. For a threat with time-to-objective $T_{\text{obj}}$, the simulation draws N samples of the full kill chain latency and computes:
Closure probability degrades under communications jamming, ROE constraints, and track quality shortfalls. Figure 1 (conceptual) shows the characteristic S-curve of $P_{\text{close}}$ vs. $T_{\text{obj}}$: near-zero for very short objectives, steep transition over the 10–25 second range, and near-certain for objectives exceeding 30 seconds. The steepness and location of this transition region are key outputs of the simulation, as they define the minimum threat timeline the system can reliably counter.
Time-critical targets (TCTs)—targets that are only vulnerable for a short window—are a primary concern for kill chain performance. The simulation classifies targets by their vulnerability window Tvuln and computes a target's "engagement margin" as:
Targets with M < 0 are not engagable under current conditions. The simulation identifies these targets and evaluates which kill chain accelerations (e.g., reducing WTA computation time, pre-positioning sensors, using cooperative engagement) would bring M above zero. This analysis directly informs JADC2 investment priorities by quantifying the marginal benefit of reducing latency in each phase.
The jadc2 phase models the integration of four major C2 systems under the JADC2 umbrella:
| System | Domain | Primary Role | Key Interface |
|---|---|---|---|
| ABMS | Air/Space | Sensor integration; data fabric | JADC2 data fabric APIs |
| C2BMC | Space/Ground | BMD command and control | Link 16, JREAP, SIPRNET |
| Aegis | Maritime | Integrated fire control | CEC, Link 16, NIFC-CA |
| THAAD | Ground | Terminal phase defense | TTT, Link 16 |
The simulation models these systems as nodes in a C2 graph, with edges representing data links and decision authorities. Each node has processing latency, decision authority scope, and connectivity constraints. The integration architecture determines which sensor-shooter pairs are possible and what intermediate decision steps are required.
Domain cross-cueing allows a detection in one domain to trigger a search or track in another. For example, a SBIRS infrared detection in the space domain can cue an Aegis radar in the maritime domain, reducing the Find and Fix phases from the radar's perspective. The simulation models cross-cueing as a directed edge in the sensor graph with an associated cueing latency and probability of successful acquisition:
Cooperative engagement extends this concept to the Engage phase. The Cooperative Engagement Capability (CEC) and Naval Integrated Fire Control-Counter Air (NIFC-CA) allow a shooter to use track data from a remote sensor that it cannot see directly. The simulation models this by relaxing the requirement that the shooter have fire-control-quality track on the target; instead, a composite track from the JADC2 data fabric can serve, provided the track quality meets engagement thresholds and the necessary C2 authorities are in place.
In a JADC2 architecture, the number of possible sensor-shooter pairs grows combinatorially. The simulation evaluates all feasible pairings and selects the optimal set using a modified WTA formulation that includes sensor-shooter latency as a cost term:
where Lij is the total kill chain latency for shooter i engaging target j via the selected sensor path, and λ is a latency penalty weight. This formulation balances kill probability against timeliness, naturally favoring low-latency pairings while still considering higher-latency options for high-value targets where longer engagement times are acceptable.
The central thesis of JADC2 is the transition from a linear kill chain to a resilient kill web. In a kill chain, each phase depends on a single predecessor; a failure at any point breaks the chain. In a kill web, each phase has multiple possible predecessors, creating redundancy:
| Attribute | Kill Chain | Kill Web |
|---|---|---|
| Topology | Linear sequence | Directed acyclic graph |
| Resilience | Single point of failure per phase | Redundant paths per phase |
| Latency | Deterministic (sum of phases) | Stochastic (shortest viable path) |
| Adaptability | Low (pre-planned) | High (dynamic re-routing) |
| Complexity | Low | High (exponential pairing space) |
The simulation models both paradigms. The kill chain mode forces a single path through the F2T2EA phases; the kill web mode evaluates all viable paths and selects the one with the best expected outcome. The difference in $P_{\text{close}}$ between the two modes quantifies the resilience benefit of JADC2. Under nominal conditions, the benefit is modest (2–5 percentage points), as the primary path is typically optimal. Under degraded conditions (jamming, sensor loss, ROE constraints), the benefit increases substantially (10–25 percentage points), as alternate paths bypass the degraded node.
The Pacific scenario models a contested maritime environment with extended ranges, limited basing, and sophisticated anti-access/area-denial (A2/AD) threats. Key characteristics:
Pacific scenarios stress the Find and Fix phases (long detection ranges), the comms phase (satellite relay latency), and the Assess phase (BDA at range). Kill chain closure times are the longest of the three theaters, making JADC2 cross-domain pairing particularly valuable.
The European scenario features a shorter-range, higher-density environment with NATO interoperability requirements:
The European theater challenges the C2 phase (ROE complexity across national boundaries) and the Track phase (dense air picture, civilian traffic discrimination). Kill chain times are shorter on average but ROE constraints create a higher variance.
The Middle East scenario models asymmetric threats in a permissive air environment:
This theater stresses the Engage and Assess phases (rapid fire doctrine, quick BDA for re-engagement decisions) and the Target phase (high target count requiring fast WTA). Kill chain times are short but the volume of engagements can overwhelm WTA processing.
Each scenario run produces a comprehensive after-action review (AAR) data set including:
All data is exported in structured JSON format for integration with external analysis tools, federation with other simulations, and archiving. The JSON schema includes scenario metadata, configuration parameters, per-timestep state snapshots, and summary statistics, enabling both real-time monitoring and post-hoc analysis.
Federation hooks allow multiple instances of the simulation to exchange track data, engagement status, and C2 decisions, supporting distributed simulation events where different teams control different domains or systems.
Validation of the simulation against published benchmarks is essential for confidence in its outputs. The validation approach follows three tiers:
Tier 1 – Component validation: Individual phase models are validated against published performance data. For example, SBIRS detection timelines are compared to the unclassified performance parameters in the Missile Defense Agency's annual reports [7], and Link 16 latency models are compared to measured latencies from open-source JTIDS/MIDS documentation [8].
Tier 2 – Kill chain validation: End-to-end kill chain timelines are compared to published engagement timelines from the Ballistic Missile Defense System (BMDS) ground tests and flight tests, as described in DoD Director, Operational Test and Evaluation reports [9]. Per-phase latency budgets are validated within 8–15% of these reference timelines, with the largest discrepancies in the Assess phase (where BDA timelines are highly scenario-dependent).
Tier 3 – Architectural validation: JADC2 integration effects (cross-cueing benefits, kill web resilience, cooperative engagement) are validated against analytical results from published studies [1, 2] and the MITRE Corporation's JADC2 modeling framework [10]. The simulation's kill web resilience improvement (10–25 percentage points under degraded conditions) is consistent with the range reported in these studies.
Known limitations include:
The c4isr-killchain-sim framework provides a comprehensive, physics-informed simulation of the complete C4ISR kill chain, from initial sensor detection through battle damage assessment. By decomposing the kill chain into seven interdependent phases—each modeled with appropriate fidelity for its dominant physics and decision processes—the simulation enables quantitative analysis of sensor-to-shooter latency, kill chain closure probability, and the operational benefits of JADC2 integration.
Three key findings emerge from the scenario analysis:
Future work includes integration with higher-fidelity network simulations (packet-level routing), human-in-the-loop C2 decision experiments, and extension to cyber and space domain kill chain phases. The modular architecture of c4isr-killchain-sim supports these extensions through its phase-graph execution model, where new phases can be added or existing phases refined without modifying the simulation framework.
© 2026 c4isr-killchain-sim · This document is a simulation research paper and contains no classified or operationally sensitive information.