Executable model
PRĀṆA–PD simulator
Patient and model settings
Objective J by arm (lower is better)
| Arm | Final mean x | Weakest domain | J total | Tracking | Burden | Safety |
|---|
Tracking = Σ(x*−x)ᵀQ(x*−x); Burden = Σuᵀ R u; Safety = λ·Σ S, where S penalises any domain below its clinical floor. Single noisy run; the Monte-Carlo summary on the paper page uses 200 seeds per cell.
Interventions chosen by the controller
| Category | Adaptive: final intensity (0–1) | Adaptive + adjunct |
|---|
Intensity is a relative research quantity, not a dose. The "medical" row is fixed by the clinician. The controller never changes it.
Domain trajectories (choose arm)
Dashed horizontal lines are the individualised targets x*. Capacity 1 = healthy reference.
Mean function across arms
Coordination reserve ρ: more is not always better
Steady state under usual care. Coupling redistributes function: higher ρ lifts the weakest domain and lowers the mean.
Controller effort over time (adaptive arm)
Dynamic coupling mask C(t): switch off a link
Untick a pair to set its coupling to 0 (a "disconnected" link) and rerun. Hypothesised pairs only.