# US 2026/0298654 A1 · Systems and methods for detecting objects and classifying zones by an autonomous vehicle using machine learning

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0298654 A1
- Document ID: US20260298654A1
- Status: Published application
- Kind: A1
- Application number: 19/705,059
- Filed: 2026-06-11
- Published: 2026-10-01
- Earliest claimed: 2024-12-20
- Assignees: Constructor Technology AG; Constructor Education and Research Genossenschaft
- Applicants: Constructor Technology AG; Constructor Education and Research Genossenschaft
- Inventors: Ilya SHIMCHIK; Ruslan MUSTAFIN; Maksim LIUBIMOV; Aleksandr BUIVAL; Serg BELL; Stanislav PROTASOV; Nikolay DOBROVOLSKIY; Laurent DEDENIS
- Web page: https://patentgrep.com/patent/US20260298654A1

## CPC

- G01C 21/3667: {Display of a road map (G01C21/3614 takes precedence; guidance using 3D or perspective road maps G01C21/3635)}
- G01C 21/3461: {Preferred or disfavoured areas, e.g. dangerous zones, toll or emission zones, intersections, manoeuvre types or segments such as motorways, toll roads or ferries}
- G01C 21/3819: {Road shape data, e.g. outline of a route}
- G01C 21/3837: {Data obtained from a single source}
- G06V 20/58: Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G01C: MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY (measuring liquid level G01F; radio navigation, determining distance or velocity by use of propagation effects, e.g. Doppler effects, propagation time, of radio waves, analogous arrangements using other waves G01S)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A system is provided for object detection and zone classification by an autonomous vehicle involves processing sensor and map data with neural networks to create Bird’s Eye View (BEV) feature vectors representing the vehicle’s surroundings. Using a detection head, objects within these vectors are identified, while a segmentation head generates a BEV segmentation map that not only highlights the detected objects but also classifies zones based on their navigability. The system compares the attributes of objects detected by both heads and updates the segmentation head’s weights accordingly. The updated segmentation head then produces a refined BEV segmentation map, which identifies at least one object and distinguishes multiple zones, including a drivable surface and an area that, while navigable, presents an elevated level of risk.

## Claims

### Claim 1

A method for detecting objects and classifying zones by an autonomous vehicle, the method comprising:
processing sensor data and map data using one or more neural networks to generate a set of Bird's Eye View (BEV) feature vectors representing an area surrounding the autonomous vehicle;
identifying, using a detection head, at least one object in the set of BEV feature vectors;
generating, using a segmentation head, a BEV segmentation map from the set of BEV feature vectors, wherein the BEV segmentation map identifies the at least one object and classifies zones around the autonomous vehicle according to respective levels of navigability;
comparing attributes of the at least one object detected using the detection head and attributes of the at least one object detected using the segmentation head;
updating weights of the segmentation head based on the comparison; and
outputting, via the updated segmentation head, an updated BEV segmentation map, wherein the updated BEV segmentation map identifies the at least one object and a plurality of zones including at least a first zone representing a drivable surface and a second zone representing an area that is navigable at an elevated level of risk.

### Claim 2 (depends on 1)

The method of claim 1, wherein the plurality of zones further includes one or more of: a current lane of the autonomous vehicle, a sidewalk, an oncoming traffic lane, and an intersection.

### Claim 3 (depends on 1)

The method of claim 1, wherein classifying zones according to respective levels of navigability comprises assigning a cost value to each zone, wherein a higher cost value corresponds to a more dangerous or prohibited area for driving.

### Claim 4 (depends on 3)

The method of claim 3, wherein the second zone representing an area that is navigable at an elevated level of risk is assigned an intermediate cost value between a cost value assigned to the first zone and a cost value assigned to a prohibited zone.

### Claim 5 (depends on 1)

The method of claim 1, wherein the one or more neural networks comprise a transformer encoder configured to convert information from a camera space to the Bird's Eye View space based on spatial relationships relative to the autonomous vehicle.

### Claim 6 (depends on 5)

The method of claim 5, wherein the transformer encoder is a spatial cross-attention transformer.

### Claim 7 (depends on 1)

The method of claim 1, wherein the sensor data comprises image data collected from a plurality of cameras operably coupled with the autonomous vehicle.

### Claim 8 (depends on 1)

The method of claim 1, further comprising generating a navigation recommendation using a path-planning model configured to process the updated BEV segmentation map.

### Claim 9 (depends on 8)

The method of claim 8, wherein the path-planning model is configured to calculate a plurality of possible paths using a cost model, and wherein generating the navigation recommendation comprises selecting a lowest cost path from among the plurality of possible paths based on a change in cost assigned by the path-planning model.

### Claim 10 (depends on 9)

The method of claim 9, wherein selecting the lowest cost path includes applying an optimizer using a Monte Carlo approximation.

### Claim 11 (depends on 8)

The method of claim 8, wherein the path-planning model is a Model Predictive Path Integral (MPPI) module.

### Claim 12 (depends on 1)

The method of claim 1, further comprising: during a training stage of the segmentation head, generating a plurality of training batches, wherein each of the plurality of training batches includes at least one sample with objects to detect.

### Claim 13 (depends on 1)

The method of claim 1, wherein attributes of the at least one object include visual features of the at least one object and a position of the at least one object.

### Claim 14 (depends on 1)

The method of claim 1, wherein comparing the attributes of the at least one object detected using the detection head and the attributes of the at least one object detected using the segmentation head comprises calculating a loss function between both attributes.

### Claim 15 (depends on 1)

The method of claim 1, wherein the at least one object is one or more of: a driving obstacle, a landmark, a pedestrian, a sign, a traffic light, another vehicle.

### Claim 16 (depends on 3)

The method of claim 3, wherein the cost value is determined based on a cost function, the cost function comprising one or more of: predicted damage to the autonomous vehicle; predicted harm to a driver and passengers of the autonomous vehicle; predicted harm to third parties; and preferences of at least one of a manufacturer and the driver regarding the cost values and associated weights.

### Claim 17

A system for detecting objects and classifying zones by an autonomous vehicle, comprising:
at least one memory;
at least one hardware processor coupled with the at least one memory and configured, individually or in combination, to:
process sensor data and map data using one or more neural networks to generate a set of Bird's Eye View (BEV) feature vectors representing an area surrounding the autonomous vehicle;
identify, using a detection head, at least one object in the set of BEV feature vectors;
generate, using a segmentation head, a BEV segmentation map from the set of BEV feature vectors, wherein the BEV segmentation map identifies the at least one object and classifies zones around the autonomous vehicle according to respective levels of navigability;
compare attributes of the at least one object detected using the detection head and attributes of the at least one object detected using the segmentation head;
update weights of the segmentation head based on the comparison; and
output, via the updated segmentation head, an updated BEV segmentation map, wherein the updated BEV segmentation map identifies the at least one object and a plurality of zones including at least a first zone representing a drivable surface and a second zone representing an area that is navigable at an elevated level of risk.

### Claim 18

A non-transitory computer readable medium storing thereon computer executable instructions for detecting objects and classifying zones by an autonomous vehicle, including instructions for:
processing sensor data and map data using one or more neural networks to generate a set of Bird's Eye View (BEV) feature vectors representing an area surrounding the autonomous vehicle;
identifying, using a detection head, at least one object in the set of BEV feature vectors;
generating, using a segmentation head, a BEV segmentation map from the set of BEV feature vectors, wherein the BEV segmentation map identifies the at least one object and classifies zones around the autonomous vehicle according to respective levels of navigability;
comparing attributes of the at least one object detected using the detection head and attributes of the at least one object detected using the segmentation head;
updating weights of the segmentation head based on the comparison; and
outputting, via the updated segmentation head, an updated BEV segmentation map, wherein the updated BEV segmentation map identifies the at least one object and a plurality of zones including at least a first zone representing a drivable surface and a second zone representing an area that is navigable at an elevated level of risk.
