An end-effector is the part of a robot that touches the world. Everything proximal — the base, the links, the joints, the arm — exists to position the end-effector, and the end-effector is the tool that does the actual job: a two-finger gripper, a multi-fingered hand, a vacuum suction array, a welding torch, a paint sprayer, or a surgical instrument. The term is precise. "Manipulator" or "arm" refers to the whole kinematic chain; "end-effector" refers specifically to the interchangeable working tool mounted at the chain's distal tip. In the patent system, end-effectors have their own home: within CPC class B25J (Manipulators), subgroup B25J 15/00 is titled "Gripping heads and other end effectors," and its descendants split out gripping by friction, by suction, by magnetic force, by interlocking fingers, and so on.

Because the end-effector is where a robot's dexterity actually lives, it is a heavily filed and strategically important part of any robotics portfolio. A company can have an excellent arm and a weak hand, or vice versa, and the patent record exposes the difference. An end-effector claim does not cover "a robot"; it covers something specific about the tool — how it grips, how it senses contact, or how its position is tracked and corrected. US10800036B1, "Tooltip stabilization," a granted patent assigned to Intrinsic Innovation LLC and classified in the B25J manipulator family, is a clean example of an end-effector control claim. It does not claim a gripper's mechanics; it claims how the tool's position is kept accurate.

receiving, from a robotic device that includes a sensor and an end effector tool, a location that is estimated for the end effector tool using sensor data that is generated by the sensor included on the robotic device; receiving, from a global sensor that is not included on the robotic device, a location that is estimated for the end effector tool— US10800036B1, claim 1, source

The claim completes the loop by "determining an adjustment to apply to the end effector tool" using both the on-board sensor's location estimate and the external global sensor's location estimate, then "providing, to the robotic device, an instruction to apply the adjustment to the end effector tool." The invention is directed to a precision problem unique to the tool tip: a robot can know its joint angles perfectly and still place its end-effector inaccurately because of flex, backlash, thermal drift, and calibration error accumulated along the arm. This claim fences off correcting the tool's position by reconciling an internal estimate with an external one. It is an end-effector claim not because it mentions a gripper's fingers but because it is directed to the accuracy of the working tool at the tip.

Reading end-effector claims for what they protect

End-effector claims fall into a few recognizable families, and identifying which family a claim belongs to is the first step in assessing it. Mechanical-grip claims protect the structure of the grasping mechanism — finger geometry, actuation linkages, compliant materials, suction-cup arrays. Sensing claims protect tactile and force feedback at the contact surface — how the tool feels what it is holding. Control and calibration claims, like US10800036B1, protect how the tool is positioned and corrected. A company filing heavily in mechanical-grip but thinly in tactile sensing has a different kind of moat than one with the reverse profile, and the difference is visible in which B25J 15/00 descendants its claims populate. Filing count alone says nothing; the distribution across these families is the signal.

It is also worth separating the end-effector from the locomotion or arm claims it travels alongside, because robotics patents often cluster. The same assignee in the same period may file an arm or locomotion patent — US9586316B1, "Determination of robotic step path" (assigned to Google LLC), claims a method of selecting a foot's "step path" for a robot "with at least three feet," which is a legged-locomotion invention, not an end-effector one — right next to an end-effector filing. Reading the independent claim is how you keep them straight: a claim reciting feet, step paths, and capture-point trajectories is about how the robot moves its body; a claim reciting an end-effector tool, gripping, or tooltip location is about what the robot does once it gets there.

Why the end-effector is the dexterity moat

The global-versus-onboard-sensor structure in US10800036B1 is itself a recurring end-effector pattern worth recognizing. The claim distinguishes "the sensor included on the robotic device" from "a global sensor that is not included on the robotic device," and computes its correction from both. That split addresses a fundamental limitation of joint-based position estimation: an arm that knows all its joint angles still cannot perfectly know its tool-tip location, because mechanical errors accumulate down the chain. An external sensor that observes the tool directly provides a ground-truth reference the arm's own proprioception cannot. The claim fences off reconciling those two sources for the end-effector specifically. Patents that recite this internal-versus-external reconciliation are claiming a calibration architecture, and they sit in a different part of the B25J landscape than claims directed to the gripper's mechanical structure or its tactile sensing. Separating these architectures is how an analyst tells a precision-calibration moat from a grasping-mechanism moat.

For an IP reader tracking the humanoid and industrial-robotics race, the end-effector claims are where defensible dexterity is won or lost. Locomotion and arm kinematics are mature; the open frontier is the hand — multi-fingered grasping, tactile sensing fine enough to handle a deformable object, and the calibration that makes sub-millimeter placement repeatable. That is why B25J 15/00 and its tactile-sensing neighbors are filling up, and why a portfolio's strength in robotics is increasingly read at the end-effector level rather than at the whole-robot level. The class B25J tells you a record concerns manipulators; the 15/00 subgroup tells you it concerns the gripping head specifically; and the independent claim tells you whether it fences off the mechanism, the sensing, or the control of the tool that does the robot's actual work.