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 › L3 · Twin Modeling & AI
L3 · Twin Modeling & AI

Receding Horizon and Real-Time Feasibility

MPC re-solves every cycle; Kronos chooses horizon length and solver so the plan is both far-sighted enough and fast enough for the control cadence.

THE STACK · click to jumpL7Ecosystem & StrategyL6Experience & VisualizationL5Applications & CopilotsL4OrchestrationL3Twin Modeling & AIL2Data FabricL1Control PlaneL0Foundation▲tlmctl▼L3 · TWIN MODELING & AIThe KRONOS-CTRL digital twin and its predictive shadow.1KRONOS-CTRL Twinlive plant state2GNNscoupled subsystems3PINNsphysics-constrained4Anomaly Ensemblesdrift & fault detection5MPCreceding-horizon control6Predictive Shadowruns seconds aheadMACHINE TIEState estimate descends to L1 control; alerts rise to L4 / L5.KRONOS FUSION ENERGYAI-NATIVE S.M.A.R.T. GENERATORTWIN MODELING & AISHEET 05REV. 2026-08L3 · AI-NATIVE STACK
L3 · Twin Modeling & AI — its place in the stack (left, click any layer) and its internal components (right). Telemetry rises; control descends.

Choosing the horizon

The prediction horizon N must be long enough to see a disturbance coming and act before a constraint binds, but every extra step costs solve time. Kronos sizes the horizon to the dominant dynamics of each control problem: shape control needs a horizon covering the coil and current-profile response; plug-density control needs to cover the ambipolar-potential response time. The horizon is expressed in twin steps, aligned to the 50-100 ms shadow so the controller sees exactly as far as the twin can faithfully predict.

Warm starting and real-time iteration

Because the problem changes only slightly cycle to cycle, Kronos warm-starts each solve from the previous solution shifted by one step. Combined with a real-time-iteration scheme, taking one Newton-type step per cycle rather than solving to full convergence, this keeps the per-cycle cost bounded and predictable, which is what a real-time controller requires.

Feasibility guarantees come from the terminal set: if each cycle can steer the horizon into a known safe terminal region, the receding-horizon scheme remains recursively feasible, it will always have a safe plan next cycle. This is the theoretical backbone that lets Kronos trust MPC to stay in the envelope indefinitely, not just for one horizon.

If a solve ever fails to complete in time, MPC does not stall: it applies the remaining valid moves from the previous plan (a feasible fallback the terminal set guarantees exists) while flagging the miss. Persistent misses trigger a controlled ramp toward a safe state. The deterministic behavior under compute pressure is designed, not incidental.

Content reviewed August 2026 · design-and-simulation stage