⌁FORMATION AUTONOMYSPACECRAFT CONTROL & RESILIENCE DIGITAL TWIN
BENCHMARK / BASELINE → PROPOSED → OPTIMIZED

Performance under pressure.

Identical conditions. Explicit trade-offs. No hidden failures.

Stress tests use dt = 0.01 s, seed 42, 60 s and 500 broadcasts per spacecraft. These are modified mission conditions, not the strict paper-parameter baseline. See the preserved baseline runs separately.

Tracking vs communication / Pareto

Tracking vs control effort / Pareto

Tracking vs modeled computation / Pareto

Normalized cost profile / lower is better

Broadcast attempts / comparison

Attitude error / selected mission

Feature ablation / combined fault

A–F / same seed and initial states

100 paired Monte Carlo scenarios / 500 runs

Full dataset ↗

Recovery is measured after the last scheduled fault transition, using a two-second terminal settling criterion. “—” means not observed within the mission. Control effort is ∫‖sat(u)‖²dt; it is not electrical energy. Computation is an operation-count model. Optimizer trained on seeds 101–102; campaign evaluated on 2026–2125.

Read engineering results ↗