Mobileye Files Patent for Building Intersection Maps From Other Cars' Drives
Every intersection is really many roads in one, since each entrance can lead to several exits. Mobileye's patent application describes a way to map each of those turns separately and then line them up so they agree.

- Publication number
- US 2026/0366886 A1(US20260366886A1)
- Applicant
- Mobileye Vision Technologies Ltd.
- Filed · Published
- Jan 29, 2026 · Nov 19, 2026
- Inventors
- Gideon WACHTEL, Yoav TAIEB, Yahel GUBERMAN, Ofer FRIDMAN, Raz Cohen MASLATON, Moshe SHENFELD, Ori SEGEL, Ofer SPRINGER
- Status when we published
- Waiting for an examiner
- Document
- 20 claims
- US classification
- 701/28
- Parent application
- Continuation of 17/824,264 (filed May 25, 2022)(17824264)
What Mobileye's intersection mapping patent does
You're stopped at a wide intersection. The car in front of you turns left, the one beside you goes straight, and you're about to turn right. Each of you takes a different path through the same patch of pavement, and a self-driving system needs to know every one of those paths.
Mobileye's filing describes a mapping system that collects short drive reports from lots of cars with cameras. Each report includes landmarks the camera spotted (a stop sign, a light pole, a crosswalk) and the route that car took. The system sorts the cars by which entrance and exit they used, boils each group down to one target path, then lines the paths up using landmarks the groups both saw.
The finished map goes out to other cars, which use it to get through the intersection. It stays small on purpose, storing paths and a few landmarks instead of full pictures.
receive drive information from each of a plurality of vehicles that traverse a road junction, the road junction including a plurality of entrances and one or more exits associated with each one of the plurality of entrances …
Translation: The system collects driving data from many different cars driving through the same intersection.
How cars' drives get sorted by entrance and exit
The patent describes a sparse map generator, a server-side system that builds lightweight maps for self-driving and driver-assist cars. It collects drive information from many vehicles that pass through the same road junction. Each report holds two things: 3D feature points (spots on objects the camera saw, each with a position in the image and a distance from the camera) and the trajectory (the path) that car followed.
The filing says that distance can come from tracking an object across several frames and using the car's own motion (ego motion, meaning how the car itself moved between frames). Then come the claimed steps:
- Sort the cars into groups by entrance-exit combination, such as everyone who came in from the south and left to the east.
- For each group, merge the points into combined 3D feature points and build a target trajectory, which a dependent claim says can be the group's average path.
- Line up the target trajectories using the combined points that show up in both groups.
- Build the sparse map with the aligned paths and send it to other vehicles.
The objects can be traffic lights, stop signs, stop lines, light poles, road markings, crosswalks, buildings or curbs. Camera position can come from GPS or from ego motion. The filing also describes versions where cars driving in opposite directions, or a rear-facing camera, help line things up.
… cause transmission of drive information for the road segment to an entity remotely-located relative to the host vehicle, wherein the drive information includes the identified at least one position descriptor associated with each of the detected one or more semantic features, the identified three-dimensional feature points, and the position information …
Translation: The car uploads its collected road data and object locations to a remote server for mapmaking.
Why crowdsourced intersection maps matter for Mobileye
If you ride in a car with driver assistance, a lot depends on the car knowing where to go in the messiest spots on the road. An intersection can offer many different routes, and a map that records each one separately could drift out of agreement with itself. This filing tackles that by lining every route up against the same landmarks.
The approach also leans on reports from ordinary cars with cameras, and the filing says keeping the map small (paths and a few landmarks instead of detailed images) cuts the cost and time of sending map data to cars. The filing names Mobileye's own EyeQ vision chips as example hardware, so the company's chips are part of the picture.
That makes this Intel's 11th filing we've tracked since September in the self-driving sensing race, a group that includes their laser sensor interference fix and Mobileye's sharper radar with fewer antennas.
Source: US 2026/0366886 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 · 73 sheets
73 drawing sheets from US 2026/0366886 A1 · click any drawing to enlarge
Show the other 52 sheets
Sheets 61 to 73 are in the full USPTO filing.

No comments yet