patent.fyi
Applications  /  US 2026/0356455
United States Patent Application Publication

US 2026/0356455 A1

Published
Nov 12, 2026
Filed
Nov 9, 2023
Earliest date on the face
Nov 18, 2022
Classification
B62M 6/50
First published here
Nov 20, 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.

1.1Rider selected speed reference

What ifWhat if a handlebar selector supplied a speed reference for each control update?
Claim language“measuring a speed of the vehicle”

A rotary selector sets a reference from 8 to 24 km/h in 2 km/h steps. The controller reads the selected reference every 20 ms and tapers motor torque as the inertial speed estimate approaches it.

1.2Ground displacement from images

What ifWhat if successive images of the road supplied displacement data for estimating speed?
Claim language“measuring accelerations and/or rotational rates of the vehicle”

A downward-facing camera captures road texture at 100 frames per second, and a range sensor supplies the camera-to-road distance. An image processor tracks the pixel displacement of road features between frames, and a trained regression model converts the displacement and distance into a forward speed estimate.

1.3Gravity corrected acceleration integration

What ifWhat if a fixed set of equations computed speed change from the inertial samples?
Claim language“by way of a machine learning model”

A complementary filter combines accelerometer and gyroscope readings to track frame pitch. The controller subtracts the gravity component along the forward axis and integrates the corrected acceleration over each 100 ms interval to obtain a speed change.

1.3Speed from steady turn geometry

What ifWhat if a geometric calculation used lateral acceleration and yaw rate to estimate speed during steady turns?
Claim language“by way of a machine learning model”

The controller rotates the inertial measurements into a level reference frame and divides horizontal lateral acceleration by yaw rate during steady turns. It updates the estimate when the absolute yaw rate exceeds 0.15 radians per second and uses the wheel speed channel between turns.

1.4Weighted speed for torque control

What ifWhat if the controller added fixed fractions of the two speed signals and used the sum to set motor torque?
Claim language“a comparison of the measured speeds and the determined speeds”

Every 20 ms, the controller multiplies the measured wheel speed by 0.8 and the inertial speed estimate by 0.2, then adds the results. A torque table uses the blended speed and measured pedal torque to set the motor current command.

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

Speed comparison for warehouse tow tractors

An electric warehouse tow tractor carries an axle encoder and a chassis-mounted inertial unit. A trained model uses windows of longitudinal acceleration and yaw rate to estimate signed changes in travel speed, while the controller calculates changes from encoder samples over the same windows. The controller reduces the traction torque limit as the difference between the two changes increases and restores the limit at a bounded rate as they converge.

Component substitution

Ground speed measurement with doppler radar

A downward-looking Doppler radar replaces the wheel speed pickup and measures longitudinal ground speed from a beam aimed forward at a fixed angle to the chassis. The controller projects the measured line-of-sight velocity onto the travel axis using the mounting angle and pitch from a chassis-mounted inertial unit. A trained model estimates speed change from acceleration and angular rate samples, and the controller reduces motor assist in proportion to the difference between that estimate and the radar speed change over the same interval.

Speed change estimation with boosted trees

A gradient-boosted decision tree ensemble replaces a recurrent neural network as the speed change estimator. The controller divides each 200-millisecond inertial window into four consecutive bins and supplies the mean acceleration and angular rate on each axis in each bin to the ensemble. The sum of the tree outputs gives the signed speed change, and the controller scales down positive motor torque as the absolute difference from the wheel speed change over that window increases.

Parameter or scale variation

Two comparison windows for torque control

A wheel encoder reports speed at 50 Hz, and a frame-mounted inertial unit samples acceleration and angular rates at 200 Hz. Every 20 milliseconds, a trained model estimates speed changes over trailing windows of 80 and 320 milliseconds, and the controller subtracts the corresponding changes in encoder speed. The controller applies a separate gain to the absolute difference for each window and uses the larger result to reduce the motor torque limit.

Combination with existing systems

Speed comparison for regenerative brake blending

An electric cargo tricycle has a rear wheel speed encoder, a chassis-mounted inertial unit, and a brake controller that blends motor regeneration with hydraulic friction braking. During braking, a trained model estimates chassis speed change from inertial samples, and the brake controller compares it with the rear wheel speed change over the same interval. When the rear wheel decelerates faster than the chassis, the controller decreases rear motor regeneration and transfers braking demand to the front friction brakes within their wheel slip limits.

Automated or data-driven variant

Automatic alignment of sensor sample times

A controller timestamps wheel speed and inertial samples with a common clock and runs a trained model on successive inertial windows to estimate speed changes. During straight coasting intervals with measurable deceleration, software tests offsets from 0 to 80 milliseconds in 5-millisecond steps and selects the offset that minimizes the median absolute difference between model and wheel speed changes. The controller applies the selected offset to subsequent wheel sample timestamps and uses the aligned comparison to adjust motor assist.

Specific version of a part

Magnetic wheel pulse measurement

A ring with 24 magnetic pole pairs rotates with the wheel hub, and a stationary Hall sensor produces one rising edge per pole pair. The controller divides the wheel circumference by 24 times the interval between consecutive rising edges to calculate speed.

Synchronized frame inertial sampling

A six-axis inertial sensor mounted to the frame measures three acceleration axes and three angular-rate axes at 200 Hz. The controller uses a common timer to timestamp inertial samples and wheel pulses, then interpolates the inertial channels to each wheel-speed measurement time.

Temporal convolution for speed change

A one-dimensional convolutional model receives a rolling window of 40 samples from the three acceleration and three angular-rate channels at 200 Hz. Training uses speed changes recorded by a ground-referenced speed sensor, and the deployed model outputs a speed change in metres per second for each 200 ms window.

Persistent speed discrepancy control

The controller subtracts inertial estimated speed from measured wheel speed at five consecutive updates spaced 20 ms apart. When all five differences exceed 1.0 m/s, it ramps the motor current limit down by 30 percent over the next 100 ms.

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.

Document AUS 7,526,402 B2
Published
Apr 28, 2009
Filed
Apr 13, 2006
Class
G01C 19/00
Document BUS 2021/0370970 A1
Published
Dec 2, 2021
Filed
Dec 29, 2020
Class
B60W 60/001
Claim 1
Document A
Document B
1.1measuring a speed of the vehicle at a plurality of points in time using a speed sensor,
The wheel tachometer is also referred to as a wheel speed sensor, and the second pose information is configured to represent the corresponding position information and/or posture information of the autonomous driving vehicle collected by the wheel tachometer, where the posture information includes speed, acceleration and other information.
B · US 2021/0370970 A1 · ¶ 58
1.2measuring accelerations and/or rotational rates of the vehicle in at least one direction, at the plurality of points in time,
A conventional IMU detects acceleration and rate of change in attitude (i.e., roll, pitch, and yaw rates) and measures these parameters over time to find the total change from the initial position.
A · US 7,526,402 B2 · ¶ 3
Generally, one IMU includes three single axis accelerometers and three single axis gyroscopes, where the accelerometer detects acceleration signal of an object on independent three-axis in a carrier coordinate system, and the gyroscope detects angular velocity signal of a carrier relative to a navigation coordinate system, so as to measure the angular velocity and the acceleration of the object in the three-dimensional space.
B · US 2021/0370970 A1 · ¶ 56
1.3determining, by way of a machine learning model, a speed and/or a speed change of the vehicle based on the measured accelerations and/or rotational rates at the plurality of points in time, and
An IMU does not detect linear velocity and position directly, which are computed parameters, by integration and double integration, based on measured accelerations, angular rates, and initial conditions.
A · US 7,526,402 B2 · ¶ 4
1.4controlling the vehicle based on a comparison of the measured speeds and the determined speeds and/or determined speed change at the plurality of points in time.
The method includes: collecting first pose information measured by an inertial measurement unit within a preset time period, and collecting second pose information measured by a wheel tachometer within the time period, where the time period is a sampling time interval when a camera collects adjacent frame images; generating positioning information according to the first pose information, the second pose information and the adjacent frame images; controlling driving of the aut
B · US 2021/0370970 A1 · ¶ 65

An empty cell means only that no passage was reproduced there.

Statements in these documents
As a closed system it is used to detect attitude and motion changes, and can be employed to determine the motion of a carrier, such as aircraft, and geo-location measurements on sensor equipment, wearable computing, e.g., as in movement detection measurements of soldiers operating on the ground, and camera stabilization platforms.
A · US 7,526,402 B2 · ¶ 3
Conventional IMUs are generally associated with high cost, large bulk, high power consumption, limited lifetime, and long turn-on time.
A · US 7,526,402 B2 · ¶ 5
However, in the implementation of the proposal, the inventor found at least the following problem: the accuracy of the estimated position and posture is low since the camera is prone to vibrate during motion.
B · US 2021/0370970 A1 · ¶ 5
A positioning method and device for accurately positioning, an autonomous driving vehicle, an electronic device and a storage medium are provided.
B · US 2021/0370970 A1 · ¶ 6
Cite aspatent.fyi, US 2026/0356455 A1, claim 1, first published 2026-11-20T01:43:50-05:00, https://patent.fyi/US20260356455/, SHA-256 d16d73eadd45f9cfbef4b8302d036d91735a24c055a3f376f3c15a5a32e339ad
First published
Permanent addresshttps://patent.fyi/US20260356455/
Content fingerprint (SHA-256)d16d73eadd45f9cfbef4b8302d036d91735a24c055a3f376f3c15a5a32e339ad
Methodv2026-11-18.2 · automated, not reviewed by a person
Correctionscorrections@patent.fyi