The assignee ranking in this dataset is led by a small group of university foundations, diabetes-device makers and continuous-glucose-monitoring specialists, with the field concentrated at the top and a long tail of single-filing entrants below them. Rather than one dominant company, the pattern looks like several parallel leaders each holding a cluster of related families, often tied to specific control algorithms or sensor-fusion approaches. Co-assignment between a university and an industry partner is one of the stronger signals in the data, suggesting some of the most defensible claims originate from academic-industry collaboration rather than purely corporate R&D.

No — filings peaked in 2017 at 81 and have declined since, with the 2022 midpoint sitting at 29 and 2026 (a partial year) far lower still. That shape is more consistent with an early period of rapid claim staking followed by consolidation than with a technology still in its growth phase. Keep in mind that the most recent one to two years understate real activity because publication typically lags filing by around 18 months, so the true 2025-2026 filing rate is higher than what is currently published.

US20140200559A1, filed by The Regents of the University of California, describes a zone model-predictive-control approach for an artificial pancreas that maintains different glucose target zones for day and night with a smooth two-hour transition between them. It matters because zone-based MPC with time-of-day modulation is a recurring architectural choice across several university-linked filers in this dataset, making this record a useful reference point for understanding what a core MPC claim looks like in this field. Anyone designing a closed-loop controller with time-varying target zones should review this family closely as part of freedom-to-operate work.

A61M (devices for introducing media into the body) and G16H (healthcare informatics) carry nearly equal weight in this dataset, at 464 and 427 records respectively, with A61B (diagnosis and monitoring) close behind at 290. G06N3 (neural network computing) and G05B (control systems) appear as smaller but present slices. Any search or clearance exercise scoped only to the traditional medical-device class will miss a substantial share of the relevant algorithmic and monitoring claims that sit in the informatics and diagnostic classes.

The dense, heavily cited core sits in closed-loop dosing and single-hormone model predictive control, but several adjacent branches show comparatively thin claim density: adaptive meal-detection algorithms, multi-hormone control combining insulin and glucagon, sensor-fusion fault-tolerance logic, personalized insulin-response model tuning, and edge-device inference for pump controllers. These sit next to the crowded core rather than inside it, meaning a new claim there is less likely to run into direct prior art from the most-cited records in this field. That said, thin density in a search corpus is not proof of a clean path — it should be paired with a targeted freedom-to-operate check before committing to a filing strategy.