This dataset tracks 200 published patent families combining AGV terminology with machine-vision-specific concepts such as visual navigation, object detection and visual servoing, filed between 2015 and mid-2026. That figure reflects one specific search definition tied to particular IPC classes, so a broader or narrower keyword set would return a different total. Because publication lags filing by roughly 18 months, the true count for the most recent one to two years will end up higher once those applications publish.

Filing activity in this space includes established logistics and robotics operators alongside a long tail of single or few-filing entrants, with citation leadership concentrated in a small number of older foundational filings. Rather than naming a fixed leaderboard here, look at the assignee ranking table on this page, which is drawn directly from the 200-family dataset and updates as new filings publish. The one clear structural signal is a coordinated co-filing pattern between Ricoh and its UK products entity.

No — filing activity peaked at 30 families in 2019 and had fallen to 13 by the 2022 midpoint, a flat-to-declining pattern rather than continued growth. Recent-year momentum figures for tracked assignees show zero filings in the latest year across the board, including a documented 100% year-on-year drop for at least one major industrial filer. Some of that recent softness is an artefact of the roughly 18-month publication lag, but the multi-year trend from 2019 onward is a genuine slowdown, not just a reporting gap.

G05D, covering control of non-electric variables, dominates with 170 of the 200 tracked records, far ahead of navigation-focused G01C (30), manipulator class B25J (27), business-process class G06Q (27) and image-recognition class G06V (24). This means the bulk of claim density sits in how the vehicle is controlled and navigated, not in the vision-detection algorithms themselves. Anyone assuming the vision layer is the most crowded part of this landscape should recheck that assumption against the IPC breakdown before drafting claims.

The clearest under-claimed ground sits in the image-recognition class G06V, which carries only 24 records against 170 in the dominant control class — proportionally thin coverage for a search built specifically around machine vision. Sub-areas such as multi-camera visual servoing fusion, load-aware visual pose correction and detection performance under adverse lighting show comparatively little dedicated claim activity in this dataset. That said, thin filing density in a class does not guarantee an open path; it means the area warrants a closer, claim-by-claim freedom-to-operate check rather than being assumed clear.

By receiving office, China leads with 81 filings against 60 for the United States, with Europe (23), WIPO/PCT (11), Japan (7) and Canada (4) well behind both. This is a receiving-office count, not a count of where assignees are headquartered, so it reflects where protection is being sought rather than where R&D originates. A team building watch lists for this technology should treat Chinese patent offices as a primary, not secondary, source.