# US 2026/0187519 A1 · Systems and methods for training a machine learning model to detect track imperfections

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

- Number: US 2026/0187519 A1
- Document ID: US20260187519A1
- Status: Published application
- Kind: A1
- Application number: 19/002,045
- Filed: 2024-12-26
- Published: 2026-07-02
- Earliest claimed: 2024-12-26
- Assignees: Constructor Technology AG; Constructor Education and Research Genossenschaft
- Applicants: Constructor Technology AG; Constructor Education and Research Genossenschaft
- Inventors: Ilya SHIMCHIK; Serg BELL; Stanislav PROTASOV; Nikolay DOBROVOLSKIY; Laurent DEDENIS
- Web page: https://patentgrep.com/patent/US20260187519A1

## CPC

- G06N 20/00: Machine learning
- A63K 1/00: Race-courses; Race-tracks
- G06N 3/045: Combinations of networks
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/08: Learning methods
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06N 3/09: Supervised learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- A63K: RACING; RIDING SPORTS; EQUIPMENT OR ACCESSORIES THEREFOR

## Abstract

A system receives scanned surface data of a physical race track. The system identifies imperfections that are not included in a virtual model of the physical race track based on the scanned surface data. The system receives sensor data capturing a first plurality of parameters of a vehicle moving along a driving path on the physical race track. The system maps, for a training dataset, at least one identified imperfection to at least one parameter in the first plurality of parameters in response to determining that the at least one parameter was captured at a location on the driving path that matches a location of the at least one identified imperfection on the physical race track. The system trains and executes a path imperfections ML model to detect track imperfections based on an input set of parameters and an input driving path using the training dataset.

## Claims

### Claim 1

A method for training a machine learning (ML) model that detects track imperfections, the method comprising:
receiving scanned surface data of a physical race track;
identifying imperfections that are not included in a virtual model of the physical race track based on the scanned surface data;
receiving sensor data capturing a first plurality of parameters of a vehicle moving along a driving path on the physical race track;
mapping, for a training dataset, at least one identified imperfection to at least one parameter in the first plurality of parameters in response to determining that the at least one parameter was captured at a location on the driving path that matches a location of the at least one identified imperfection on the physical race track;
training a path imperfections ML model to detect track imperfections based on an input set of parameters and an input driving path using the training dataset; and
executing the path imperfections ML model.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
performing a racing simulation in which a virtual vehicle moves along a simulated driving path on the virtual model, wherein the simulated driving path corresponds to the driving path;
generating virtual sensor data capturing a second plurality of physical parameters of the virtual vehicle moving along the simulated driving path;
detecting at least one difference between the first plurality of parameters and the second plurality of parameters at a first set of points in the driving path;
mapping, for the training dataset, the at least one identified imperfection to the at least one difference; and
training the path imperfections ML model to detect the track imperfections based on an input set of parameter differences and an input driving path using the training dataset.

### Claim 3 (depends on 1)

The method of claim 1, wherein the virtual model comprises a plurality of points each with track information, and the driving path comprises a subset of the points where the vehicle has driven, and wherein the path imperfections ML model modifies track information for a first set of points in the subset of the points, wherein the first set of points are associated with locations of the track imperfections.

### Claim 4 (depends on 3)

The method of claim 3, wherein the path imperfections ML model is further trained to output an updated driving path with the track imperfections.

### Claim 5 (depends on 3)

The method of claim 3, wherein the track information is indicative of one or more of location, elevation, surface type, obstruction, and damage.

### Claim 6 (depends on 3)

The method of claim 3, further comprising:
executing a generative ML model that receives, as an input, the track imperfections indicated in the subset of the points and the virtual race track, and outputs an updated virtual race track with possible imperfections across other points of the plurality of points that are not in the subset.

### Claim 7 (depends on 6)

The method of claim 6, wherein the generative ML model is trained using an optimization algorithm that compares the updated virtual rack track with the scanned surface data of the physical track.

### Claim 8 (depends on 1)

The method of claim 1, wherein the first plurality of parameters comprises one or more of: vertical acceleration, roll, pitch, yaw, point cloud data acquired by at least one sensor, and suspension travel.

### Claim 9 (depends on 8)

The method of claim 8, wherein the at least one sensor includes one or more of: a LiDAR radar, a camera, a stereo camera, and a depth sensor.

### Claim 10 (depends on 8)

The method of claim 8, wherein the first plurality of parameters further comprises one or more of: lateral acceleration, longitudinal acceleration, tire pressure, vehicle internal temperature, wheel speed data, global positioning system (GPS) data, track surface temperature, vibration data, humidity, ambient temperature, optical sensor data.

### Claim 11 (depends on 1)

The method of claim 1, wherein training the path imperfections ML model comprises updating weights of the path imperfections ML model to reduce a difference between predicted track imperfections and target track imperfections.

### Claim 12

A system for training a machine learning (ML) model that detects track imperfections, comprising:
at least one memory; and
at least one hardware processor coupled with the at least one memory and configured, individually or in combination, to:
receive scanned surface data of a physical race track;
identify imperfections that are not included in a virtual model of the physical race track based on the scanned surface data;
receive sensor data capturing a first plurality of parameters of a vehicle moving along a driving path on the physical race track;
map, for a training dataset, at least one identified imperfection to at least one parameter in the first plurality of parameters in response to determining that the at least one parameter was captured at a location on the driving path that matches a location of the at least one identified imperfection on the physical race track;
train a path imperfections ML model to detect track imperfections based on an input set of parameters and an input driving path using the training dataset; and
execute the path imperfections ML model.

### Claim 13 (depends on 12)

The system of claim 12, wherein the at least one hardware processor is further configured to:
perform a racing simulation in which a virtual vehicle moves along a simulated driving path on the virtual model, wherein the simulated driving path corresponds to the driving path;
generate virtual sensor data capturing a second plurality of physical parameters of the virtual vehicle moving along the simulated driving path;
detect at least one difference between the first plurality of parameters and the second plurality of parameters at a first set of points in the driving path;
map, for the training dataset, the at least one identified imperfection to the at least one difference; and
train the path imperfections ML model to detect the track imperfections based on an input set of parameter differences and an input driving path using the training dataset.

### Claim 14 (depends on 12)

The system of claim 12, wherein the virtual model comprises a plurality of points each with track information, and the driving path comprises a subset of the points where the vehicle has driven, and wherein the path imperfections ML model modifies track information for a first set of points in the subset of the points, wherein the first set of points are associated with locations of the track imperfections.

### Claim 15 (depends on 14)

The system of claim 14, wherein the path imperfections ML model is further trained to output an updated driving path with the track imperfections.

### Claim 16 (depends on 14)

The system of claim 14, wherein the track information is indicative of one or more of location, elevation, surface type, obstruction, and damage.

### Claim 17 (depends on 14)

The system of claim 14, wherein the at least one hardware processor is further configured to:
execute a generative ML model that receives, as an input, the track imperfections indicated in the subset of the points and the virtual race track, and outputs an updated virtual race track with possible imperfections across other points of the plurality of points that are not in the subset.

### Claim 18 (depends on 17)

The system of claim 17, wherein the generative ML model is trained using an optimization algorithm that compares the updated virtual rack track with the scanned surface data of the physical track.

### Claim 19 (depends on 12)

The system of claim 12, wherein the first plurality of parameters comprises one or more of: vertical acceleration, roll, pitch, yaw, point cloud data acquired by at least one sensor, and suspension travel.

### Claim 20

A non-transitory computer readable medium storing thereon computer executable instructions for training a machine learning (ML) model that detects track imperfections, including instructions for:
receiving scanned surface data of a physical race track;
identifying imperfections that are not included in a virtual model of the physical race track based on the scanned surface data;
receiving sensor data capturing a first plurality of parameters of a vehicle moving along a driving path on the physical race track;
mapping, for a training dataset, at least one identified imperfection to at least one parameter in the first plurality of parameters in response to determining that the at least one parameter was captured at a location on the driving path that matches a location of the at least one identified imperfection on the physical race track;
training a path imperfections ML model to detect track imperfections based on an input set of parameters and an input driving path using the training dataset; and
executing the path imperfections ML model.
