A New Intel Self-Driving Tech Patent Uses Other Drivers' Paths to Find Where Lanes Split and Merge
Mobileye's patent application teaches a mapping system to spot where a road forks or merges by watching where real drivers peel apart or come together. It then checks the guess against a camera image from that exact spot.

- Publication number
- US 2026/0366885 A1(US20260366885A1)
- Applicant
- Mobileye Vision Technologies Ltd.
- Filed · Published
- Dec 5, 2025 · Nov 19, 2026
- Inventors
- Dori SHAPIRA, Daniel BRAUNSTEIN, Yoram GDALYAHU, Kfir VIENTE, Maxim SCHWARTZ, Eli PEKER, David HUBERMAN, Moshe LEVIN, Jonathan BARLEV, Ora ZACKAY, Adi HAYAT, Andras FERENCZ, Dor EAGELBERG
- Status when we published
- Waiting for an examiner
- Document
- 25 claims
- US classification
- 701/533
- Examiner
- SMITH, JORDAN T (Art Unit 3666)
- Parent application
- Continuation of 17/294,984 (filed May 18, 2021)(17294984)
What Mobileye's lane-fork mapping actually does
Ever hovered at a highway fork, unsure which lane carries you where? A human driver reads the paint and the signs. A self-driving system leans on a map, and that map has to know that one lane turns into two.
Mobileye's idea is to build that knowledge from other drivers. Cars that have already driven a stretch of road report the paths they took. When two cars start in the same lane and then drift apart by more than a set distance, the system suspects the road forks. It then pulls a camera image from that spot and looks for lane paint or road structures that back up the suspicion. The image also helps pin down exactly where the fork begins.
Once confirmed, the map stores one path that carries on down one branch and a second path that splits off from it. The same trick runs in reverse for merges, where two lanes funnel into one.
… determine a divergence between the first actual trajectory and the second actual trajectory based on comparison of a lateral distance between the first actual trajectory and the second actual trajectory to threshold distance; …
Translation: The system finds where cars start drifting apart by measuring the sideways gap between their paths.
How two drivers' paths reveal a lane split
The system runs on a server that collects driving records from many vehicles. Each record includes location points a car logged while driving a road segment, which gives the server each car's actual trajectory (the real path it took). Cameras on those cars can supply images, or the server can pull them from storage.
Then comes the fork-finding routine:
- Find two trajectories that share a first lane and then end up in two different lanes.
- Measure the lateral distance (the sideways gap) between them and compare it to a threshold.
- Pick the image taken where they part, and analyze it for a lane mark or road structure that supports a split.
- Use that image to refine where the split starts.
Finally the server updates its road navigation model (the map data cars drive by). One target trajectory (the ideal path) covers the lane before the split and continues down one branch. A second target trajectory branches off it into the other lane. Dependent claims also have the server remove anomalies, meaning gaps between paths that are not real splits.
The merge version mirrors this: a convergence between two paths, confirmed by an image, produces two target trajectories that join. The description adds that a host car can receive this model and choose which branch to follow based on what it sees around it.
… update a vehicle road navigation model to include a first target trajectory and a second target trajectory that branches from the first target trajectory after the lane split feature.
Translation: The maps used by self-driving cars are updated to show the exact paths drivers take when a road divides.
Why lane forks and merges matter for car maps
Lane forks and merges are the spots where a driving system most needs to know the plan ahead of time. A map that already says "this lane becomes two" gives the car's software a clear path through them, and your car does not have to figure it out from paint alone.
The approach also saves effort. Mapping crews do not have to survey every interchange. The cars already on the road supply the evidence, and the image check screens out false alarms such as a car that simply wandered across a lane line. The description ties this to Mobileye's sparse map, which stores roads as compact paths instead of heavy imagery, and names its EyeQ chips as possible in-car hardware.
This is the 12th Intel filing we've tracked since September in the self-driving sensing race, following earlier work on building intersection maps and a sharper car radar.
Source: US 2026/0366885 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.
Get our take in your Top Stories
Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Intel patent news breaks.
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings · 68 sheets
68 drawing sheets from US 2026/0366885 A1 · click any drawing to enlarge
Show the other 52 sheets
Sheets 61 to 68 are in the full USPTO filing.

No comments yet