Sony · Filed Sep 18, 2025 · Published Sep 10, 2026 · verified — real USPTO data

Sony Patents a Robot Arm That Knows When to Push and When to Give Way

Programming a robot arm to handle every possible obstacle is a nightmare, so Sony is trying a different approach: let the arm be stubborn in one direction and physically yielding in another, with no extra instructions required.

A robot arm with a hand portion interacts with a sliding door, demonstrating its ability to apply force in specific directions. Drawing from patent filing US 2026/0264237 A1.
A robot arm with a hand portion interacts with a sliding door, demonstrating its ability to apply force in specific directions.
See all 20 drawings from this filing ↓
Publication number US 2026/0264237 A1
Applicant SONY GROUP CORPORATION
Filing date Sep 18, 2025
Publication date Sep 10, 2026
Inventors YASUHIRO MATSUDA
CPC classification 700/245
Grant likelihood Medium
Examiner KHAYER, SOHANA T (Art Unit 3657)
Status Docketed New Case - Ready for Examination (Jun 18, 2026)
Parent application is a National Stage Entry of PCTJP2024011294 (filed 2024-03-22)
Document 20 claims

How Sony's robot arm handles surprise obstacles

Ever tried to move a piece of furniture through a doorway you weren't quite sure was wide enough? You push forward confidently, but let the sofa rotate and flex sideways so it doesn't jam. Sony's patent describes robot arms that work on the same principle.

The arm gets told which direction to actively push toward a goal, and which directions to stay physically relaxed in. If something unexpected gets in the way from a relaxed direction, the arm bends to it rather than forcing through or stopping dead. You don't have to pre-program every possible collision scenario.

The idea is to let robots complete tasks even when their environment isn't perfectly mapped or controlled. Instead of needing a detailed script for every situation, the arm can feel its way through, succeeding more often without more information.

From the filing · CLAIM 1
a setting unit configured to set a movement direction in which at least one part of the manipulator is caused to move and a passive direction in which the part has passivity …

Translation: The system defines which way the robot arm should actively push and which way it should limp or yield.

How the manipulator splits motion from compliance

The patent describes a manipulator (the technical term for a robotic arm or similar device) built around two complementary concepts: a movement direction and a passive direction.

The movement direction is where the arm is actively trying to go. A control unit applies force and tracks position along this axis, driving the arm toward its target. The passive direction is perpendicular to that (or at some other defined angle) and here the arm is deliberately made compliant, meaning it offers little resistance. If an object pushes the arm off-course in the passive direction, the arm yields rather than fighting back.

A setting unit within the system configures both directions before a task begins. The claim is deliberately broad: it covers any part of the manipulator, not just the end effector (the hand or tool at the tip), so compliance could apply to individual joints or whole arm segments.

  • Set the active (push) direction for the goal
  • Set the passive (yield) direction for expected disturbances
  • Let the control unit manage motion using both parameters together

The result is a robot that can complete tasks even when its surroundings are imprecise or partially unknown, without needing a detailed model of every object it might touch.

From the filing · THE ABSTRACT
To enable execution of a task with a high success rate without detailed information.

Translation: The goal is to help the robot succeed at jobs even when it lacks precise data about its surroundings.

What yielding robot arms mean for factories and homes

The biggest practical barrier to robots doing useful work outside of tightly controlled factory lines is that the real world is unpredictable. A robot that stops or breaks whenever something is slightly out of place is expensive to run. A robot that can absorb small surprises physically, without new software, is a lot more deployable.

Sony's steady investment in robot control points toward applications in consumer and service robotics, where environments are far messier than a car factory. If this approach works as described, it could reduce the setup time needed to put a robot arm to work in a warehouse, a kitchen, or a care facility, making those deployments cheaper and less fragile for whoever is running them.

This is the seventh Sony filing we've tracked since July, with earlier entries covering touch and distance sensing and dodging obstacles faster in our robot grasping and moving watchlist.

Editorial take

Factory robots are brittle in a very specific, very expensive way: they fail the moment the physical world doesn't match their expectations. A part arrives at a slight angle, a tray shifts a few millimeters, and the line stops. At industrial scale, that fragility adds up to enormous amounts of lost time and human intervention.

Sony's patent addresses that failure mode directly, by designing an arm that can sense resistance and give way in certain directions while still pressing forward in others, rather than demanding that everything be perfectly placed before it acts.

Whether a control concept like this translates into a robot that handles genuine disorder reliably is an open question. But the underlying problem is real, widespread, and costly enough that meaningful progress on it would matter well beyond any single factory floor.

There are more where this came from

We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.

The drawings

20 drawing sheets from US 2026/0264237 A1 · click any drawing to enlarge

Patent filing page

Source. Full patent text and figures from the official USPTO publication PDF.