Skip to content
Technology How it works Breeder — Hyperion Burner — Aegis Burner — MetroVolt AI-Native Architecture Magnets Fuel cycle Safety Roadmap
Solutions AI & Data Centers Defense & Government Grid & Baseload Neutron Detection Quantum
Learn Technical Library
Proof Publications Whitepapers Technical Library Open Science & Reproducibility The Honest Gates
Company About / Mission Leadership Environment Health & Safety Investors Careers Press Contact
3D Model
AI Architecture › Resiliency & Operations
Resiliency & Operations

Control Loop Latency Budget

Decomposing the real-time path into sense, transport, infer, decide, actuate - and the fallback taken when any stage overruns.

STRATEGY / SLOW ▲ ▼ MICROSECOND REAL-TIMEL7Ecosystem & Strategytelemetry ▲ control ▼open ▸L6Experience & Visualizationtelemetry ▲ control ▼open ▸L5Applications & Copilotstelemetry ▲ control ▼open ▸L4Orchestrationtelemetry ▲ control ▼open ▸L3Twin Modeling & AItelemetry ▲ control ▼open ▸L2Data Fabrictelemetry ▲ control ▼open ▸L1Control Planetelemetry ▲ control ▼open ▸L0Foundationtelemetry ▲ control ▼open ▸PHYSICAL S.M.A.R.T. GENERATOR PLANTBREEDER · HYPERION1R0 1.2 m · A 2.5 · 16.84 T · δ −0.30BURNER · TANDEM MIRROR2317 T throat · 26.49 T plug · fₙ 5.44% · DEC1 center stack + plasma · 2 high-field plug · 3 expander → direct converterCOLOR GRAMMAR strategy AI-workflow infra/data models reactor/DECLINE SEMANTICStelemetry (µs)controlKRONOS FUSION ENERGYAI-NATIVE S.M.A.R.T. GENERATORMASTER BLUEPRINTSHEET 01REV. 2026-08L0-L7 · 2 MACHINES
The AI-Native S.M.A.R.T. Generator Master Blueprint — eight layers (L0→L7), one control stack, wired to both machines. Telemetry rises in microseconds; control descends the same path.

The five stages

Every hard-real-time decision on either machine traverses the same five stages. Budgeting means assigning each stage a worst-case execution time and proving the sum plus jitter clears the deadline. The stages are: sense (diagnostic acquisition and digitization), transport (deterministic fabric to the control node), infer (state estimate or surrogate model evaluation), decide (controller law), and actuate (power electronics response).

Determinism over average speed

Resiliency demands bounded worst case, not fast average. A model that is quick 99% of the time but occasionally stalls is unusable in a 1 ms loop. We therefore forbid dynamic memory allocation, unbounded loops, and non-deterministic branches in the fast path, and we pin the inference model to a fixed-operation form.

python
# Overrun guard: if inference exceeds its slice, fall back to the last-good
# controller law rather than block the actuation deadline.
def fast_step(state, budget_us):
    t0 = now_us()
    try:
        u = surrogate_infer(state, deadline_us=t0 + budget_us)
    except DeadlineExceeded:
        u = last_good_law(state)      # deterministic PID fallback
        flag('infer_overrun')
    actuate(u)
    return u

Jitter and margin

Jitter (variation in loop period) is as dangerous as latency. We size a 2x margin between worst-case path and deadline so that a single slow cycle does not cascade. Persistent overruns are logged for postmortem and can trigger a controlled derate. The full timing hierarchy is in the timing model.

On the breeder the tightest budget is vertical-position control of the elongated delta -0.30 plasma; on the burner it is plug-field regulation near the 26.49 T operating point where the coil is already stressed 3-3.9x at the design bore, so the actuation authority is deliberately conservative.

Content reviewed August 2026 · design-and-simulation stage