The leading assignee in this dataset holds 24 of the 663 records in scope, with the fifth-ranked assignee at 14 and the tenth-ranked at 12. That gentle decline rather than a sharp cliff means the top 5 combined account for only 14.3% of all records and the top 10 for 24.0% — this is not a field controlled by one or two dominant portfolios. A freedom-to-operate review should therefore look across the ranked leaders rather than assume one company's patents cover the field.

Filings rose from 24 in 2017 to a peak of 98 in 2025, passing through 37 at the 2022 midpoint, which is real growth over the period as a whole. However, several previously active filers show no filings in the latest tracked year, and the 2026 figure is partial because publication typically lags filing by around 18 months. Read together, the evidence points to a field that has already passed its fastest growth phase rather than one still accelerating.

The largest IPC category is G06F, electric digital data processing, covering 29.3% of the 663 records, followed by G05B control and regulating systems at 21.0% and G06Q business and administrative data processing at 18.3%. Smaller but notable shares include G06N AI-based computing at 11.3% and H04L digital transmission at 9.2%. Because a single record can carry multiple IPC classes, these figures add up to more than 100% of the record base and should be read as overlapping coverage, not mutually exclusive segments.

US20140330402A1, filed by AspenTech Corporation and published in 2014, claims an MPC controller that operates across configuration, identification and model-adaptation modes, using user-specified model structure information to build and later adapt a working master model. It predates the bulk of the 663-record dataset, so it functions more as a reference point for how model-structure claims were framed before the 2017–2025 filing surge than as a live blocking concern for every new filing. Anyone drafting around model-adaptation or model-structure claims in this space should still review it directly, since its priority date is early relative to most of the field.

The evidence points to several under-claimed branches relative to the size of the core G05B and G06F categories: constraint back-off calibration under live plant-model mismatch, benefit-attribution methods that quantify value split between the RTO and MPC layers, and fallback control logic that governs what happens when the optimizer and controller disagree. These are narrower than the dominant classification categories but sit directly on the terms used in the search scope. A first claim in any of these areas should specify the triggering condition (mismatch threshold, disagreement event, or attribution method) precisely, since that specificity is what will distinguish it from the broader, more heavily filed control and optimization claims already in the corpus.