patent.fyi
Applications  /  US 2026/0356508
United States Patent Application Publication

US 2026/0356508 A1

Published
Nov 12, 2026
Filed
May 9, 2025
Earliest date on the face
May 9, 2025
Classification
B64G 99/00
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.

1.1Tracking data drives propagation

What ifWhat if a regression model forecasts the orbit from timestamped position and velocity measurements without receiving body masses?
Claim language“masses of bodies”

A processor fits a local polynomial to each Cartesian position component using a sliding window of position and velocity measurements. It evaluates the fitted derivatives at future times and locates inflections from sign changes in planar curvature.

1.2Projecting acceleration onto the normal

What ifWhat if the processor projects acceleration onto the trajectory normal and forecasts its zero crossings?
Claim language“computing a curvature value”

For planar motion at nonzero speed, the processor rotates the unit velocity vector by 90 degrees and projects acceleration onto that normal. It interpolates between consecutive propagated samples with opposite signs to estimate the time of an inflection.

1.2Testing sampled turn direction

What ifWhat if the processor classifies the turn direction of successive position triples using a geometric orientation test?
Claim language“computing a curvature value”

The processor forms two consecutive displacement vectors from each triple of planar position samples and takes the sign of their determinant. When adjacent triples have opposite signs, it reports the middle timestamps of those triples as an estimated interval for the change in bending direction.

1.3Solving the forecast interval together

What ifWhat if a solver computes the entire forecast interval in one coupled trajectory solution?
Claim language“applying a recursive formula”

A collocation solver fits position and velocity polynomials over the forecast interval and imposes the equations of motion at fixed nodes. The processor derives curvature from the fitted velocity and acceleration at each requested time.

1.3Integrating a curvature rate

What ifWhat if each update adds a curvature increment derived from a curvature rate?
Claim language“a nonlinear scaling function”

A planar dynamics model supplies velocity, acceleration, and jerk to compute the time derivative of signed curvature. The processor multiplies that derivative by the step duration and adds the increment to the current curvature.

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

Curvature forecasting for tethered underwater vehicles

A topside computer receives the underwater vehicle position, velocity, and wet mass along with the tether payout and fairlead position. It projects the vehicle track onto the horizontal plane, computes signed curvature from the current velocity and acceleration, and uses the horizontal distance to the fairlead as the radius input. At each forecast step it multiplies the latest curvature by a nonlinear factor calibrated from maneuver records spanning the operating range of tether payout. It reports a predicted inflection interval when consecutive forecast samples have opposite signs.

Component substitution

Cubic spline for curvature scaling

A piecewise cubic spline replaces the radius-dependent scaling function. The spline input is the current separation divided by a reference separation, and its knot values are fitted to successive measured curvature ratios from samples whose starting curvature exceeds the measurement noise floor in magnitude. At runtime the predictor clips the normalized separation to the fitted interval, evaluates the spline, and multiplies the current signed curvature by the result. The product becomes the curvature forecast for the next time step.

Phase plane inputs for curvature scaling

A phase-plane calculation uses separation and separation rate in place of a radius-only input to the scaling function. It centers separation on a reference separation, divides the two signals by their respective oscillation amplitudes, and computes phase with the two-argument arctangent. A periodic lookup table supplies a signed scaling factor from the phase and uses linear interpolation between adjacent bins, including across the wrap point. The predictor multiplies the current signed curvature by the interpolated factor to produce the next forecast sample.

Parameter or scale variation

Separate forecast intervals for two trajectories

A predictor for a pair of coasting spacecraft maintains a separate signed curvature and forecast clock for each trajectory. It derives the two initial curvatures from the position and velocity estimates and the modeled accelerations, using the body masses in the acceleration model. It advances the first curvature every 0.5 seconds for 10 seconds and the second every 2 seconds for 60 seconds, using a scaling function calibrated for each step duration. Each forecast sample carries its own timestamp so the two sequences align at their shared sample times.

Combination with existing systems

Curvature forecasts for ground antenna scheduling

A ground tracking scheduler connects the curvature predictor to its station visibility calculator and antenna mount model. The predictor receives the spacecraft state and the modeled body masses, projects the trajectory into the station line-of-sight tangent plane, and recursively forecasts signed angular curvature in that plane using slant range as the radius input. The mount model combines the forecast curvature with the predicted angular speed to estimate the acceleration normal to the line-of-sight track. The scheduler flags pass segments where the corresponding mount acceleration exceeds the configured limit.

Automated or data-driven variant

Adaptive fitting from tracking measurements

A state estimator combines range, range-rate, and angle measurements with a dynamics model to produce positions, velocities, and accelerations for computing signed curvature. The predictor expresses its scaling factor as a weighted sum of nonlinear basis functions of normalized range and range rate. Recursive least squares updates the weights by fitting each measured curvature to the preceding measured curvature multiplied by that step's factor, with measurement variance used to weight the residual. The fitted factor supplies the next forecast, and a new curvature estimate initializes the recursion after each measurement update.

Specific version of a part

Reporting an inflection time window

The processor compares the signs of successive internally forecast signed curvature samples. It sends the timestamps that bracket each sign reversal to a guidance scheduler.

Time tagged cartesian state records

The input contains a common epoch and one record per body with three position components in meters, three velocity components in meters per second, and mass in kilograms. The processor reads the position and velocity components in a shared inertial frame.

Signed curvature from state derivatives

For a planar trajectory, the processor takes the determinant of the current velocity and acceleration vectors. It divides the determinant by the cube of the speed to obtain signed curvature in inverse meters.

Timestamped curvature output record

After each recursive update, the processor creates a record containing the target time t + Δt and the predicted curvature in inverse meters. It writes the record to a circular buffer that a display or guidance process reads in timestamp order.

Cite aspatent.fyi, US 2026/0356508 A1, claim 1, first published 2026-11-20T03:03:17-05:00, https://patent.fyi/US20260356508/, SHA-256 28a3f27c7ea745dba7add56e32d183b3c45e99321290b2b3ba02bae798623f3b
First published
Permanent addresshttps://patent.fyi/US20260356508/
Content fingerprint (SHA-256)28a3f27c7ea745dba7add56e32d183b3c45e99321290b2b3ba02bae798623f3b
Methodv2026-11-18.2 · automated, not reviewed by a person
Correctionscorrections@patent.fyi