The world's first patent
for controlling what
must not change.
When a disruption hits your production line, conventional systems rebuild the entire schedule. COGNIROBO's patented Control Layer does not. It identifies what is already committed — and recalculates only what is safe to change.
What was obvious on the shop floor
had never been formalized as a system.
Every experienced production manager knows intuitively: don't touch what's already running. But no system had ever encoded that constraint as a structured, patentable control mechanism — until COGNIROBO filed in 2023.
"A control mechanism that determines which production instructions must remain unchanged and which can be adjusted — classifying all operations into Frozen, Postponed, and Rescheduled groups — so that only the strictly necessary portion of the schedule is ever recalculated."
Existing systems in 2023 asked: "How do we find the optimal schedule?" This patent asks first: "What must we protect before any optimization begins?"
A disruption triggers a full rebuild.
Everything is recalculated from zero.
- Entire schedule thrown away and recomputed
- Already-prepared instructions are discarded
- Workers receive completely new task sequences
- WIP, staged materials, and setup state are ignored
- Recovery takes hours — not minutes
A disruption triggers a targeted fix.
Only what must change is touched.
- Already-committed instructions remain frozen
- Blocked operations are classified and held
- Only the impacted zone is recalculated
- Shop-floor state is a first-class constraint
- Real-time recovery aligned with actual conditions
Three groups. One control principle.
Every production instruction is classified into exactly one of three groups at the moment a disruption is detected. Only the third group is ever recalculated.
Must Not Change
Instructions already committed to the floor. Protected unconditionally from recalculation.
- Scheduled to start within 30 minutes
- Changeover / setup already completed
- Materials already loaded or staged
- Priority or high-value customer orders
Cannot Execute Now
Operations blocked by a real-world constraint. Held without cancellation — ready to resume.
- Material shortage or supplier delay
- Equipment failure or scheduled maintenance
- Quality deviation or regulatory hold
- Operator or resource unavailability
Can Be Adjusted
The only group that is recalculated — targeted, minimal, and fast.
- Downstream operations not yet started
- Flexible capacity windows exist
- No committed setup or material loaded
- No frozen precedence dependency
Disruption Detected
Equipment failure, material shortage, quality hold — any real-time event triggers the Control Layer.
Instructions Classified
Every pending operation is evaluated against shop-floor state and sorted into Frozen, Postponed, or Rescheduled.
Targeted Recalculation
AI optimization runs only on the Rescheduled group — leaving all Frozen instructions untouched.
Stable Recovery
Updated instructions reach the floor immediately. Workers and machines continue without a full stop.
Different industries. Same control principle.
The Control Layer applies wherever production instructions carry real-world commitment — and where rebuilding the entire schedule would cause more damage than the original disruption.
High-value WIP must never be disrupted.
- Freeze lots already in process (Frozen)
- Postpone affected downstream lots
- Reschedule only unstarted operations
- Maintain throughput and yield stability
Synchronized flows cannot be stopped mid-cycle.
- Preserve fixed assembly / welding / painting ops
- Postpone only impacted parts or supplier delays
- Reschedule downstream without stopping the line
- Avoid full line stoppage
Validated processes cannot be changed.
- Keep all validated steps fixed (Frozen)
- Postpone batches with deviations or holds
- Reschedule only compliant operations
- Maintain GMP compliance and traceability
"This seems obvious —
so why is it patentable?"
A fair question. Here is the honest answer.
The world focused on math, not people.
For decades, production scheduling research asked one question: "How do we find the mathematically optimal schedule?" Every major system — MRP, APS, AI optimizers — was built around total optimization: minimize cost, maximize throughput, find the best solution globally.
No one asked the opposite question first.
Human constraints were never formalized.
On every shop floor, workers and supervisors had always known:
- "Don't touch the line that's already running."
- "The changeover is done — that job stays."
- "We told the operator — don't change it now."
- "Only adjust the part that's actually broken."
This was tribal knowledge. Real. Practiced every day. But never encoded as a system.
That gap is exactly what the patent covers.
COGNIROBO formalized what operators already knew — and built it into a structured, repeatable control mechanism: Dynamic Fixing Condition.
The novelty was not a new algorithm. It was the first time anyone defined, in software, a method to separate "what must not change" from "what can be adjusted" — using real-time shop-floor state as the deciding factor.
That combination — human operational constraint + AI recalculation scope — had never been systematized or patented anywhere. Until 2023.
Why the US Patent Office granted it.
The USPTO evaluates novelty and non-obviousness. The prior art search confirmed: no existing patent or publication described a scheduling system that:
- Dynamically classifies instructions as Frozen, Postponed, or Rescheduled
- Uses shop-floor operational state (not just time or priority) as the fixing condition
- Limits AI/optimization scope to only the Rescheduled group
- Preserves already-committed instructions unconditionally
The idea felt obvious on the floor. As a defined, replicable system — it was new.
- Mathematical optimization of the full schedule
- AI finds the globally best solution
- Human constraints treated as soft inputs
- Shop-floor state ignored or averaged
- Disruption = rebuild from scratch
- Human operational constraints defined first
- Shop-floor state is a hard constraint
- Frozen instructions protected unconditionally
- AI applied only where it is safe
- Disruption = targeted fix of affected zone only
"Conventional optimization asks: how do we find the best solution?
This patent asks first: which instructions must be protected from any change — and applies AI only where it is safe and necessary to do so."
Dynamic Fixing Condition (動的固定条件) · Control Layer Patent · COGNIROBO Inc. · 2023
Ready to evaluate the Control Layer?
We provide API access, interactive demos, and integration support for qualified manufacturers, system integrators, universities, and research partners.