Granted Patent B2

US 12,731,407 · App. 18/348,590 · Granted Sep 8, 2026

Inventors: Magnus Gyllenhammar (Pixbo, SE); Adam Tonderski (Västra Frölunda, SE)

Assignee: ZENSEACT AB

G06V20/56G06V10/7792

Abstract

Claims (38)

1 . A computer-implemented method for updating a perception function of a vehicle having an Automated Driving System (ADS) having a self-supervised machine-learning algorithm configured to generate an output based on one or more ingested images and a machine-learning algorithm for an in-vehicle perception module trained to provide one or more in-vehicle perception functions based on one or more ingested images, the method comprising:

obtaining one or more images of a scene in a surrounding environment of the vehicle;

updating one or more model parameters of the self-supervised machine-learning algorithm in accordance with a self-supervised machine learning process based on the obtained one or more images;

generating a first output by processing the obtained one or more images by using the self-supervised machine-learning algorithm;

generating a supervisory signal for a supervised learning process based on the first output; and

updating one or more model parameters of the machine-learning algorithm for the in-vehicle perception module based on the obtained one or more images and the generated supervisory signal in accordance with the supervised learning process.

2 . The method according to claim 1 , wherein the generated supervisory signal comprises the generated first output, and wherein the obtained one or more images and the generated supervisory signal forms training data for the machine-learning algorithm for the in-vehicle perception module.

3 . The method according to claim 1 , wherein the generating of the supervisory signal comprises processing the first output by using a secondary machine learning algorithm trained to generate a second output based on the generated first output;

wherein the supervisory signal comprises the second output;

wherein the obtained one or more images and the supervisory signal forms training data for the machine-learning algorithm for the in-vehicle perception module.

4 . The method according to claim 3 , wherein the second output comprises at least one of object classification, depth estimation, bounding box, segmentation mask, and object trajectory.

5 . The method according to claim 1 , further comprising:

detecting anomalous image data by using a machine-learning classification system trained to distinguish new experiences from experiences known to the self-supervised machine-learning algorithm in the obtained one or more images and to output an anomaly value;

adding a weight to the supervisory signal based on the anomaly value.

6 . The method according to claim 5 , wherein the machine-learning classification system comprises an autoencoder trained on the same dataset as the self-supervised machine-learning algorithm, and wherein the anomaly value is a reconstruction error.

7 . The method according to claim 1 , further comprising:

transmitting the updated one or more model parameters of the self-supervised machine-learning algorithm and the updated one or more model parameters of the machine-learning algorithm for the in-vehicle perception module to a remote entity;

receiving a set of globally updated one or more model parameters of the self-supervised machine-learning algorithm from the remote entity, wherein the set of globally updated one or more model parameters of the self-supervised machine-learning algorithm are based on information obtained from a plurality of vehicles comprising a corresponding self-supervised machine-learning algorithm;

receiving a set of globally updated one or more model parameters of the machine-learning algorithm for the in-vehicle perception module from the remote entity, wherein the set of globally updated one or more model parameters of the machine-learning algorithm for the in-vehicle perception module are based on information obtained from a plurality of vehicles comprising a corresponding machine-learning algorithm for the in-vehicle perception module;

updating the self-supervised machine-learning algorithm based on the received set of globally updated one or more model parameters of the self-supervised machine-learning algorithm; and

updating the machine-learning algorithm for the in-vehicle perception module based on the received set of globally updated one or more model parameters of the machine-learning algorithm for the in-vehicle perception module.

8 . The method according to claim 1 , wherein the self-supervised machine-learning algorithm is a Masked Autoencoder (MAE).

9 . The method according to claim 1 , wherein the one or more in-vehicle perception functions comprises at least one of:

a semantic segmentation function, an instance segmentation function, an object classification function, an object detection function, a free-space estimation function, and a tracking function, an object trajectory prediction function.

10 . A non-transitory computer-readable storage medium comprising instructions which, when executed by a computing device, causes the computer to carry out the method according to claim 1 .

11 . A system for updating a perception function of a vehicle having an Automated Driving System (ADS) having a self-supervised machine-learning algorithm configured to generate an output based on one or more ingested images and a machine-learning algorithm for an in-vehicle perception module trained to provide one or more perception functions based on one or more ingested images, the system comprising control circuitry configured to:

obtain one or more images of a scene in a surrounding environment of the vehicle;

update one or more model parameters of the self-supervised machine-learning algorithm in accordance with a self-supervised machine learning process based on the obtained one or more images;

generate a first output by processing the obtained one or more images by using the self-supervised machine-learning algorithm;

generate a supervisory signal for a supervised learning process based on the first output; and

update one or more model parameters of the machine-learning algorithm for the perception module based on the obtained one or more images and the generated supervisory signal in accordance with the supervised learning process.

12 . The system according to claim 11 , wherein the control circuitry is further configured to:

detect anomalous image data by using a machine-learning classification system trained to distinguish new experiences from experiences known to the self-supervised machine-learning algorithm in the obtained one or more images and to output an anomaly value; and

adding a weight to the supervisory signal based on the anomaly value.

13 . The system according to claim 12 , wherein the machine-learning classification system comprises an autoencoder trained on the same dataset as the self-supervised machine-learning algorithm, and wherein the anomaly value is a reconstruction error.

14 . A vehicle comprising:

one or more sensors configured to capture images of a scene in a surrounding environment of the vehicle; and

a system according to claim 11 .

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (32)