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.
No entries.
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
Lane selection for yard tractors
Yard tractors record their speed, steering angle, and position on a map of parallel travel lanes in a freight terminal. A simulator replays each recorded approach to a slower tractor and generates a lane change for a recorded lane hold or a lane hold for a recorded lane change. A lane selection model trains on the paired runs using time to the assigned bay and minimum clearance from surrounding vehicles as training targets.
Component substitution
Traffic response using car following
An intelligent driver model replaces fixed trajectory replay for vehicles in the source and destination lanes. The simulator initializes each surrounding vehicle from its recorded position and speed, then computes its acceleration from the gap and closing speed to its current leader. As the assisted vehicle crosses the lane boundary in the simulated run, the simulator updates leader assignments and supplies the resulting gaps and speeds to the lane change training process.
Parameter or scale variation
Paired simulations across maneuver durations
The training pipeline resamples recorded vehicle tracks at 20 hertz and initializes an 8-second simulation from each selected point in a lane-hold recording. For each initial state, the simulator generates lane changes lasting 3, 4, and 5 seconds using lateral trajectories with zero lateral velocity and acceleration at the endpoints. The training pipeline pairs all three runs with the same recorded lane hold and uses minimum gap, peak lateral acceleration, and longitudinal progress as training targets.
Combination with existing systems
Lane assistance with adaptive cruise control
The simulator runs an adaptive cruise controller in the longitudinal control loop for both the lane-hold and lane-change runs. The controller starts both runs with the recorded set speed and following time gap, then updates throttle and brake commands from the distance and relative speed to the selected lead vehicle. The lane change model trains on the paired results using travel progress, minimum following gap, and peak braking demand.
Automated or data-driven variant
Automatic pairing from vehicle logs
A batch processor aligns timestamped vehicle pose and surrounding-object tracks with mapped lane boundaries. It labels a segment as a lane change when the vehicle center crosses into an adjacent lane and stays there for at least 2 seconds, and labels a segment as a lane hold when the center remains in the same lane throughout. For each accepted segment, the processor initializes the simulator before the maneuver or at a matched point in a lane-hold segment, requests the opposite maneuver, and stores both trajectories with a shared training identifier.
Simulation sampling by model disagreement
An ensemble of lane change models scores recorded situations using vehicle speed, lane geometry, and front and rear gaps in the adjacent lane. A sampler ranks situations by the variance of the ensemble's lane change probabilities and sends the highest-ranked situations to a simulator. The simulator generates the opposite maneuver for each recording, and the training process adds the paired results with clearance and progress targets to the next training batch.
Cite aspatent.fyi, US 2026/0356369 A1, claim 1, first published 2026-11-20T00:27:46-05:00, https://patent.fyi/US20260356369/, SHA-256 3647a074893c7157aa0e62543df1dd39836b55f05647626770796b67deaeff0a