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.
1.1Grasp geometry from live depth
What ifWhat if the controller calculates contact points from live depth measurements for each pick?
Claim language“storing data relating to a plurality of holding patterns”
An overhead depth camera measures the exposed surfaces of the parts in the bin. The controller locates two opposing surfaces within the gripper’s jaw travel and calculates the wrist pose and jaw opening from their measured positions.
1.2Fingers conform during closure
What ifWhat if a mechanical differential distributes finger travel as the gripper closes around the workpiece?
Claim language“selecting one holding pattern”
A depth sensor locates the top of the workpiece, and the robot lowers an open three-finger gripper over it. A cable differential transfers travel to the free fingers as each finger meets the surface, and a motor current limit caps the closing force.
1.3Switch sequenced pick motion
What ifWhat if limit switches and a vacuum switch govern robot motion at a fixed pickup station?
Claim language“controlling the robot based on the one holding pattern”
A vibratory feeder delivers one workpiece at a time to a pickup nest beneath a pneumatic Cartesian robot. A lower limit switch triggers suction, a vacuum switch triggers the lift, and an upper limit switch triggers horizontal travel to the discharge station.
1.4Candidates from surface geometry
What ifWhat if a candidate generator derives contact locations from measured surface geometry?
Claim language“based on the holding pattern”
The generator searches a depth map for flat patches wider than the suction cup and fits a plane to each patch. It places the cup center at the patch center and aligns the approach with the fitted surface normal to produce one candidate per patch.
1.5Automatic acceptance after trials
What ifWhat if repeated lift trials determine which candidates enter the holding pattern library?
Claim language“receiving a user instruction”
The controller executes ten lift trials for each candidate and measures vacuum pressure and workpiece displacement during each hold. It adds a candidate to the library when all ten trials maintain a pressure differential of at least 40 kPa and keep displacement below 2 mm.
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
Linen sorting with vacuum pickup
A robot above a laundry sorting belt carries a segmented vacuum head that lifts individual linen items from a loose pile. The controller selects a stored holding pattern that specifies the head orientation, active suction zones, and vacuum dwell time. It generates candidate patterns by changing the active zones and dwell time while retaining the pickup pose. An operator selects a candidate for storage, and the controller makes that pattern available for later pickups.
Component substitution
Nearest neighbor selection from measured dimensions
A k-d tree nearest-neighbor search replaces the pattern selector, and each tree entry links a pair of measured workpiece dimensions to a stored holding pattern. Before pickup, a depth sensor measures the visible length and width of the target workpiece, and the controller queries the tree using those measurements. The controller executes the returned pattern and generates candidates by changing its jaw closing speed and force limit. When an operator adds a candidate, the controller records the dimensions measured during its trial and inserts the new entry in the tree.
Case-based generation of holding pattern candidates
A case-based reasoning algorithm replaces the candidate generator and reads a table of previous successful pickups indexed by gripper type, measured part width, and measured part mass. It retrieves cases with the same gripper type and ranks them by distance in normalized width and mass. For each of the three closest cases, it copies the recorded jaw force and lift speed into a candidate that retains the selected pattern's pickup pose. The operator reviews the candidates, and the controller adds each selected candidate to the holding pattern store.
Parameter or scale variation
Holding patterns for lightweight molded caps
A bulk feeder presents molded caps 12 to 20 mm in diameter to a robot with a parallel-jaw gripper fitted with elastomer pads. The controller selects a stored pattern and generates six candidates by pairing closing forces of 1, 2, and 3 N with closing speeds of 5 and 10 mm/s. Each candidate retains the selected pattern's pickup pose and pre-grasp jaw opening. The operator selects a candidate after a test pickup, and the controller stores its force and speed settings with the retained pose and opening.
Combination with existing systems
Holding patterns linked to production recipes
A production recipe manager sends the robot controller a part identifier and a gripper configuration identifier at each job change. The controller uses these identifiers to retrieve a bank of holding patterns and selects a pattern from that bank for bulk pickup. It generates candidates by varying the jaw opening and lift acceleration of the selected pattern. The operator selects a candidate at the cell terminal, and the controller stores it under the active identifiers for retrieval when the same recipe returns.
Automated or data-driven variant
Grip trials measured by tactile sensors
A parallel-jaw gripper carries tactile arrays on both jaw pads and reports contact force and movement of the contact patch during lifting. The controller creates candidate force and lift-acceleration settings from the selected holding pattern and runs each candidate on a series of test picks. It ranks the candidates by the fraction of lifts that stay below a preset contact-motion threshold and displays peak contact force alongside each result. An operator selects a candidate from the trial results, and the controller saves its settings as a holding pattern.
Specific version of a part
Workpiece indexed grasp records
Nonvolatile flash memory holds a table keyed by workpiece type and pattern number. Each record contains a workpiece-relative wrist pose, a jaw opening in millimeters, and a closing force in newtons.
Selection by measured clearance
The selector aligns each stored pattern for the workpiece type with the pose estimated from a depth image. It rejects patterns whose modeled finger paths intersect neighboring parts and selects the remaining pattern with the greatest minimum clearance.
Pattern guided force control
The controller transforms the selected wrist pose and approach vector from workpiece coordinates into robot base coordinates. At the target pose, a load cell measures jaw force, and a feedback loop regulates it to the stored force setpoint during the lift.
Candidates from wrist offsets
The generator copies an existing pattern and creates candidates with wrist rotations of 5 and 10 degrees in both directions about the approach axis. Each candidate retains the original tool center point, jaw opening, and closing force.
Section 3 · quoted from USPTO records
Other work
Passages reproduced word for word from two earlier-published U.S. patent documents, placed beside the language of claim 1. No characterization is offered. Each passage is marked A or B for the document it comes from; ¶ numbers are the paragraph numbers in the USPTO’s published full text of that document.
1.1a storage unit for storing data relating to a plurality of holding patterns for holding the workpiece with an end effector of the robot;
The analysis of the workpiece 310 (that is, for a particular part design) to generate the set G of robust grasp locations is performed once at the outset, and then used in the grasping and motion planning calculations for each individual workpiece in a bin.
B · US 2021/0308866 A1 · ¶ 30
1.2a selection unit for selecting one holding pattern of the plurality of holding patterns for the workpiece;
The developer may extend the bin picking software to include new sensor support, new grippers, new pose estimation (Matcher) algorithms, new boundary generators, and new grasp script selectors.
A · US 2021/0023710 A1 · ¶ 59
A grasp planning method for an industrial robot, said method comprising: determining a set of robust grasps and a set of stable poses for a workpiece shape; determining an initial pose of a workpiece; providing a goal pose for the workpiece; removing infeasible grasps from the set of robust grasps, selecting a feasible grasp, and planning robot motions to grasp and move the workpiece directly from the initial pose to the goal pose; when no feasible grasps exist, selecting an
B · US 2021/0308866 A1 · claim 1
1.3a control unit for controlling the robot based on the one holding pattern;
In some embodiments, robotic bin picking process 10 may include picking up the workpiece using the robot and/or end effector. This may include specifying one or more poses of the robot relative to the pose of the workpiece.
A · US 2021/0023710 A1 · ¶ 74
At box 620, robot motion commands are provided from the controller to the robot, causing the robot to physically grasp and move the workpiece from the initial pose A to the goal pose B. The grasping and movement of the workpiece from the initial pose A to the goal pose B is direct if possible (from the box 618), or uses one or more intermediate state M otherwise (from the boxes 614 and 616).
B · US 2021/0308866 A1 · ¶ 50
1.4a creation unit for creating a plurality of holding pattern candidates based on the holding pattern;
1.5a reception unit for receiving a user instruction for adding at least one of the created holding pattern candidates as the holding pattern; and
1.6a storage control unit for causing the storage unit to store the received holding pattern candidate as the holding pattern.
An empty cell means only that no passage was reproduced there.
Statements in these documents
However, robotic bin picking is a particularly difficult task to master as the amount of accuracy and precision required is often beyond the capabilities of the system.
A · US 2021/0023710 A1 · ¶ 2
The method may further include displaying, at a graphical user interface, the pose of the workpiece relative to the robot to allow for visual verification.
A · US 2021/0023710 A1 · ¶ 4
In light of the circumstances described above, there is a need for a robot grasp planning technique which adaptively identifies the most efficient part gripping and motion sequence.
B · US 2021/0308866 A1 · ¶ 5
Each individual workpiece in the bin is evaluated to identity a set of feasible grasps, and the workpiece is moved to the final position and pose if such a direct movement is possible.
B · US 2021/0308866 A1 · ¶ 6
Cite aspatent.fyi, US 2026/0355957 A1, claim 1, first published 2026-11-19T17:33:15-05:00, https://patent.fyi/US20260355957/, SHA-256 aeda24d33bcf71b2b862538af2dd3957360803d79f2285765087b9ebc337307f