Control that learns the plant — no model supplied.

The engine identifies your plant online and drives it with an SDRE controller in about a millisecond per step. Delivered as a $99/mo cloud layer — supervisory: we compute setpoints, your PLC executes. Private and edge deployments — under separate contract.

Pricing: Layer 3 — $99/mo; everything in Filtration, plus device control; we compute setpoints, your PLC executes. Full pricing →

Talk to us See the benchmarks
Matches the oracle ~1 ms per step MIMO with hidden channels

How it works

One identified model, two ways to spend it.

No plant model is supplied. The engine identifies the nonlinear plant online from the stream, then the identified model is consumed two ways: an SDRE controller for millisecond-budget real-time loops, or MPC/NMPC on the same model where a hard constraint must never be violated — trading compute for guarantees.

01 · NO-LAG

Channel-independent output

SDRE-class per-channel control computed in real time from the cleaned signal — about a millisecond per step on benchmark plants.

02 · NMPC

Latency-agnostic

Optimization on your full compiled offline model, with built-in stability protections — for loops where quality matters more than lag.

03 · MIMO

Correlated groups

Whole groups of channels computed jointly — all cross-correlations honoured; mix grouped and independent channels freely.

See Pricing for how the layers fit together.

Honest boundaries

Supervisory, civil, not safety-certified.

Cloud control is supervisory: we compute setpoints, your PLC executes. Fast and safety-critical loops run on your own edge under private contract. Civil applications only; not a safety-certified controller. Benchmarks above are representative v17 core results on canonical plants — benchmarked, not field-proven.

See Twin Ops