patent.fyi
Applications  /  US 2026/0354337
United States Patent Application Publication

US 2026/0354337 A1

Published
Nov 12, 2026
Filed
May 3, 2024
Earliest date on the face
May 25, 2023
Classification
A01J 5/0175
First published here
Nov 18, 2026

Generated automatically. Not reviewed by a person. Not legal advice or an opinion.

Section 1 · machine-generated

What if?

Questions about how individual parts of claim 1 might be changed. A question may describe something different from what the claim language recites, or a version of it; no view is offered on which.

No entries.

Section 2 · machine-generated

Variations

Further arrangements of the same subject matter, described here as of the publication date at the foot of this page.

Other field of use

Sequential cup placement for dairy goats

A goat milking stall uses an ear tag reader, an upward-facing depth camera, and a robot arm that picks up two teat cups from a rack. The controller reads the goat identifier, locates both teats, and attaches the cups one at a time in the stored order for that goat. An inline vacuum sensor confirms each seal, and the controller records the time from the start of the first approach until both cups seal. The controller compares the median times for the two orders over repeated visits and stores the order with the shorter median time.

Component substitution

Running averages for attachment order selection

A controller uses exponentially weighted running averages in place of a trained model to select among four stored attachment orders. It maintains a separate average for each cow and order, using elapsed attachment time plus a fixed time penalty for each retry as the score for a visit. It tries each order once during the first four visits and updates the corresponding average after every visit. It normally selects the order with the lowest average and rotates through the other three orders on every tenth visit to refresh their scores.

Parameter or scale variation

Timed retries for four cup attachment

A robot arm attaches four cups in the order stored for the identified cow and starts a 4-second timer at the beginning of each approach. A pressure sensor in each cup line confirms attachment when the vacuum remains within its configured operating band for 300 milliseconds. If the timer expires, the arm withdraws the cup and advances to the next teat, then makes one additional attempt at each missed teat after the first pass. The controller records total elapsed time and missed cups for each tested order and selects the order with the shortest mean time among orders with the highest observed completion rate.

Combination with existing systems

Attachment control with quarter milk meters

Each teat cup connects through its own milk hose to a quarter milk meter, with a pressure sensor mounted on the cup side of the meter. The robot controller stores the teat position assigned to each sensor and meter channel when it attaches a cup. The controller timestamps the pressure-confirmed seal and the first sustained milk flow on each channel. For each cow, it ranks the tested orders by attachment completion rate, then by median time from the first cup approach until all four channels register flow.

Automated or data-driven variant

Order selection from measured teat geometry

A depth camera below the udder measures the positions of the four teat tips in the robot coordinate frame before the first cup approach. The controller logs the cow identifier, measured positions, attachment order, total attachment time, and number of retries for each visit. Two gradient-boosted regression models use these records to predict time and retries for each of the 24 possible orders. The controller commands the order with the lowest predicted time after adding a fixed time penalty per predicted retry.

Attachment order updates after cow movement

A depth camera tracks the teat tips while vacuum sensors report which cups are sealed. After each seal confirmation, the controller refreshes the positions of the unattached teats and enumerates the orders that remain. A regression model trained on the cow’s recorded approach times and retries estimates the remaining attachment time for each order using current teat positions and the arm pose. The arm moves to the first teat in the order with the shortest estimated time, and the controller repeats the calculation after the next seal.

Cite aspatent.fyi, US 2026/0354337 A1, claim 1, first published 2026-11-18T21:18:11-05:00, https://patent.fyi/US20260354337/, SHA-256 399ff3f3555b84c2f5d730195dc469018551f2407880e979d8c3b697afe4306a
First published
Permanent addresshttps://patent.fyi/US20260354337/
Content fingerprint (SHA-256)399ff3f3555b84c2f5d730195dc469018551f2407880e979d8c3b697afe4306a
Methodv2026-11-18.2 · automated, not reviewed by a person
Correctionscorrections@patent.fyi