A soft sensor (also called a virtual or inferential sensor) is a model that estimates a process variable from other measured signals rather than measuring it directly. In this dataset, the scope is narrowed to soft sensors paired with mechanisms for keeping the underlying model current, such as online learning, drift detection, recursive estimation, transfer learning or explicit update-interval control. That combination is what separates this landscape from a broader soft-sensor search: the emphasis is on how the model adapts after deployment, not just on the initial estimation method.

The ranking covers 38 assignees, and it is genuinely top-heavy: the top five hold 57.5% of the 160 records in scope, and the top ten hold 74.4%. The leading assignee alone accounts for 26 records, well ahead of the fifth-place figure of 9. Beyond that top tier, though, a long tail of companies files in smaller numbers, spanning industrial control, AI infrastructure and process-specific specialists, so the field is not a two-company duopoly.

Filings grew from 32 in 2021 to a peak of 37 in 2024, a 16% increase, and 2024 is the most recent year with a reasonably complete publication record. The apparent drop-off in 2025 and 2026 is a data artefact: patent publication typically lags filing by around 18 months, so recent years are always undercounted until later filings work through the pipeline. Reading the 2025-26 dip as a real slowdown would be a mistake based on this dataset alone.