COA Decision Support Agent Dashboard

An interactive, zero-dependency HTML5/CSS3/Vanilla JavaScript operational research and decision support platform for autonomous Courses of Action (COA) selection, dynamic replanning, and resilience analysis.

---

Quick Start (How to Launch)

No installation, build tools, npm, or server required! You can run the dashboard in any of the following ways:

• Direct Browser Launch: Double click or open index.html in Microsoft Edge, Google Chrome, Firefox, or Safari.
• Local HTTP Server (Optional): Run Start-Process "index.html" from PowerShell or your preferred local web server.

---

Directory Structure

coa-decision-support-dashboard/
├── index.html              # Main multi-tab dashboard layout
├── README.md               # User guide & operational documentation
├── css/
│   └── style.css           # Modern defense & operations research theme
└── js/
    ├── charts.js           # Zero-dependency SVG charting utility
    ├── paper1_engine.js    # BDI reasoning engine
    ├── paper2_engine.js    # Degraded conditions, sweeps, Monte Carlo, Sobol
    └── app.js              # UI controller, events, and reactivity

---

Core Operational Modules

Module 1: BDI Dynamic Planning Agent

• Dual-Plan Single-Resource (BDI Cycles):
  • Models consecutive plans P1 and P2 sharing budget Total1 = 80.
  • Generates all feasible pairs, scores via 0.7 × util + 0.3 × coverage.
  • Executes the 4-phase BDI event sequence:
    • Cycle 1: Commits to COA1 + COA5 (R1 = 75, score = 0.96).
    • Cycle 2 (Resource Shock): Logistics shortfall cuts budget to 65. Abandonment and replan commits to COA1 + COA6 (R1 = 65, score = 1.00).
    • Cycle 3 (COA Failure): COA1 route compromised and removed. Replanning commits to COA3 + COA4 (R1 = 65, score = 1.00).
    • Cycle 4 (New Intel): Unconditional trigger forces re-deliberation; maintains COA3 + COA4.
  • Displays full live terminal execution log.
• Probabilistic Mission Reliability (Monte Carlo):
  • Enhances COAs with success probabilities p.
  • Surviving pair COA3 + COA4 (P1 = 0.80, P2 = 0.95 ⇒ P_joint = 0.760).
  • Executes N=1000 (or custom) Monte Carlo trials, plotting convergence to empirical success rate (≈ 0.767).
• Dual-Resource Multi-Objective Optimization:
  • Evaluates dual constraints (Total1 = 7, Total2 = 8) and multi-resource weights (w1 = 1.0, w2 = 2.0).
  • Explains why pairs involving P2_Y are pruned (R1 = 8 > 7).
  • Demonstrates why P1_B + P2_X (score = 16.0) beats P1_A + P2_X (score = 21.0) because Resource 2 is penalized twice as heavily.
• High-Dimensional Combinatorial Space (5 Plans × 5 Resources):
  • Combinatorial evaluation of all 5⁵ = 3125 plan combinations across 5 resource dimensions.
  • Identifies optimal combination: P1_C4, P2_C0, P3_C0, P4_C3, P5_C0 with combined resources [10.62, 7.77, 8.90, 7.65, 9.96] and score 67.98.
  • Resource usage bar chart against uniform budget of 15.0.

---

Module 2: COA Evaluation in Degraded Environments

• Interactive COA Candidate Settings:
  • All candidate COAs are fully editable via interactive table inputs:
    • Required Resource r_i
    • Baseline Effectiveness b_i
    • Degradation Robustness ρ_i
  • Real-time reactivity: Any edit immediately updates scores, threshold detection, continuous crossover curves, and sensitivity sweeps.
  • Dedicated "Reset COAs" button to quickly restore baseline settings.
• Sequential Degradation & Threshold Shift:
  • Implements the mathematical scoring model:

    δ = (u + d_c + d_a) / 3
    E_i = max(0, 1 - r_i / R_total),   S_i = ρ_i · (1 - δ)
    V_i = α · E_i + (1 - α) · S_i,   Φ_i = b_i · (1 - δ) · V_i

  • Tracks degradation across: Initial (δ=0) → Minor comms delay (δ=0.113) → Conflicting intel (δ=0.411) → Severe interference (δ=0.845).
  • Detects and flags the Threshold Shift from COA4 to COA1.
• Resource Constraint Sensitivity Sweep:
  • Sweeps R1 ∈ [40, 80] showing the S-shaped phase transition and crossover at R1 ≈ 68.
  • Interactive SVG plot showing selection frequencies.
• Viability Weight (α) Sensitivity Sweep:
  • Evaluates α ∈ [0.25, 0.75] with R1 = 70.
  • Shows steeper crossover between α = 0.375 and α = 0.500.
• Continuous Degradation & Crossover Curves:
  • Continuous trajectory of all COA scores as degradation varies from 0 to 1.0.
  • Interactive sliders for R1 and α with analytical crossover calculation.
• Global Multi-Parameter Sensitivity (Monte Carlo):
  • Uniform sampling across the 3D threat cube (u, d_c, d_a) ∈ [0, 1]³.
  • COA4 win rate (≈ 98.34%), COA1 win rate (≈ 1.66%), and Spearman rank correlation (≈ -0.57).
• Sobol Global Variance Decomposition:
  • First-order (S1 ≈ 0.329) and Total-order (ST ≈ 0.337) indices.
  • Verifies equal tripartite distribution and negligible second-order interactions (S2 ≈ 0.0039).