1.1Fixed seated exercise station
A motor pivots the handle through a programmed arc while the trainee sits beside the frame. A load cell between the grip and the pivot arm measures push and pull during the movement.
Generated automatically. Not reviewed by a person. Not legal advice or an opinion.
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.
A motor pivots the handle through a programmed arc while the trainee sits beside the frame. A load cell between the grip and the pivot arm measures push and pull during the movement.
Two synchronized cameras track markers on the thigh and shank while the robot drives the knee. The controller calculates knee angle from the marker positions and records the angular deviation from the programmed trajectory.
A low-pass filter smooths the signed force signal. The controller multiplies the filtered value by a fixed gain and adds the result to the programmed joint speed, subject to upper and lower speed limits.
Surface electrodes over the elbow flexor and extensor muscles record electrical activity. The controller compares the root-mean-square signal from each muscle with its calibrated threshold to select a flexion, extension, rest, or simultaneous activation state.
A rotary encoder measures joint angle and supplies samples for estimating angular velocity. The controller calculates commanded motor torque from the angle and velocity errors relative to the programmed trajectory.
Further arrangements of the same subject matter, described here as of the publication date at the foot of this page.
An assembly worker wears a powered shoulder joint mounted between a torso harness and an upper-arm cuff. A bidirectional load cell between the cuff and the driven link measures tangential interaction force while the actuator supplies lift, and a joint encoder measures angular velocity. The controller classifies the force as assisting motion, opposing motion, or sustained during a hold using force sign and a velocity deadband. It increases lift during upward effort, reduces lift during downward effort, and maintains bounded support during a hold.
A torsion spring between the gearbox output and the joint output link replaces a direct interaction force transducer. An encoder on each side of the spring measures its deflection, which the controller multiplies by the calibrated spring stiffness to obtain transmitted torque. The controller subtracts the modeled torque for gravity, friction, and acceleration of the output link and divides the residual by the cuff moment arm to estimate tangential interaction force. A force deadband and the measured motion direction determine whether the controller holds, increases, or reduces motor torque.
A double-acting pneumatic cylinder replaces the electric drive at the wearable knee hinge and connects to the hinge through a crank. A load cell at the shank cuff measures interaction force, and a hinge encoder supplies motion direction and the current cylinder moment arm. The controller combines force magnitude and direction with hinge motion to classify the interaction as assisting, opposing, or inside a deadband and selects a corresponding torque command. A pressure feedback loop drives proportional valves to achieve the cylinder force required by the torque command while enforcing configurable pressure and joint angle limits.
A wearable arm trainer has independently driven elbow and wrist joints, each with an encoder and an interface load cell positioned at a known moment arm. The controller samples both channels at 200 hertz and computes a rolling mean interaction torque over 100 milliseconds for each joint. In a bench configuration, the elbow and wrist channels use classification deadbands of 0.5 and 0.15 newton meters, respectively, and actuator torque limits of 4 and 1 newton meters. Every 20 milliseconds, each channel increases assistance for input in the direction of motion, decreases assistance for opposing input, and holds the current command for input inside the deadband.
A wearable knee trainer combines a powered hinge, a shank-cuff load cell, a joint encoder, and a surface electrical stimulation unit with electrodes over the knee extensor muscles. During a cued extension, the controller uses cuff force and joint velocity to classify interaction as extension effort, resistance, or low effort. Low effort during the extension cue triggers a stimulation burst within configured amplitude and duration limits, and the motor torque controller uses force samples between bursts to update assistance. Resistance above a configured threshold ends the burst and ramps down motor assistance.
A powered elbow orthosis logs cuff force, joint angle, joint velocity, and motor current while a trainee performs cued assistance, resistance, and relaxation trials. Software computes mean force, force slope, and force–velocity correlation over overlapping 150-millisecond windows and uses the trial cues to label the training data. A trained decision tree classifies each live window, and the controller accepts a class change after three consecutive windows agree. A lookup table maps the accepted class to a bounded torque increment, torque decrement, or hold command.
A current regulator maintains the preset motor current during each exercise stroke. The audio circuit plays a low tone when measured force opposes travel and a high tone when it acts in the direction of travel.
A strain-gauged aluminum flexure connects the forearm cuff to the driven elbow link. A bridge amplifier converts flexure strain into a signal that the controller samples at 200 hertz and converts to tangential force using a stored calibration.
The controller assigns the filtered force a positive sign when it acts in the direction of joint travel. A reading above 5 newtons for 100 milliseconds produces an assisting classification, and a reading below minus 5 newtons for the same duration produces a resisting classification. Readings within the threshold band produce a neutral classification.
The controller fits a straight line to force samples in a rolling 50-millisecond window and uses the slope as the force rise rate. Separate amplitude and slope thresholds distinguish a sustained load from a rapid force transient.
The controller selects a motor current setpoint from a table indexed by the force classification. It ramps the setpoint downward for sustained resistance and upward for sustained assistance within a configured joint torque limit.
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.
Robotic rehabilitation system 102 includes one or more robotic motion machines 110, a control system 112 for controlling motors of the robotic motion machines 110, and one or more sensors 114 for collecting user interaction data while one or more users 104 perform physical rehabilitation training using the robotic motion machines 110.A · US 2018/0318648 A1 · ¶ 13
The initial input forces f1 and f2 (at least one of which might be in part if not totally derived from human operator for example (and at least one of which might be in part if not totally derived from a signal from the coupling device via the respective actuator)) are fed to the coupling controller 16 via an input port into a computational array comprising a number of features such as first block functions 17a and 17b, and node 18.B · US 11,400,590 B2 · ¶ 30
In some examples, analyzing the user interaction data includes determining, using comparison data one or more of: error data, rate of learning data, visuospatial performance data, working memory performance data, attention deficit data and force data.A · US 2018/0318648 A1 · ¶ 16
In some examples, control system 112 is configured for providing assistive and resistive haptic feedback to user 104 by controlling motors of the robotic motion machines. Control command producer 128 can send commands to alter the assistive and resistive haptic feedback.A · US 2018/0318648 A1 · ¶ 22
This patent application describes the use of (modified) Lagrange multipliers to calculate a virtual force which is then added to demanded force within the admittance control loop of each actuator, resulting in an optimal coupling. Use of modified Lagrange multipliers allows for simulation of coupled actuators, solve initial condition problems, fading of coupling forces, and simulating break out conditions where actuators are coupled up till a certain force level.B · US 11,400,590 B2 · ¶ 8
An empty cell means only that no passage was reproduced there.
Robotic physical rehabilitation systems must have versatility to compete with traditional therapies in providing assessment and treatment for disability resulting from a diversity of diagnoses.A · US 2018/0318648 A1 · ¶ 3
The methods and systems can, in some examples, help diagnose and quantify the level of impairment of the patient—and at a much faster rate than is feasible than through some conventional clinical means.A · US 2018/0318648 A1 · ¶ 4
Limitations of current technology include that they are only valid for the specific case when initial conditions of x1(t) and x2(t) are equal, only provide specific solution for the constraint x1(t)=x2(t), and or do not allow for specialized coupling behavior such as simulating break out behavior where coupling force between both controls is capped.B · US 11,400,590 B2 · ¶ 4
Use of modified Lagrange multipliers allows for simulation of coupled actuators, solve initial condition problems, fading of coupling forces, and simulating break out conditions where actuators are coupled up till a certain force level.B · US 11,400,590 B2 · ¶ 8
patent.fyi, US 2026/0355036 A1, claim 1, first published 2026-11-19T07:16:45-05:00, https://patent.fyi/US20260355036/, SHA-256 912373ca7776ccd7ade4634cdbd4542582582e405c1104421cc047002e0d7ada