Method for training a central artificial intelligence module

Patent metadata

Field	Value
Publication number	US-11586203-B2
Application number	US-201816162527-A
Country	US
Kind code	B2
Filing date	Oct 17, 2018
Priority date	Oct 30, 2017
Publication date	Feb 21, 2023
Grant date	Feb 21, 2023

Abstract

A method for training a central artificial intelligence module (“AI module”) for highly or fully automated operation of a vehicle, the central AI module to translate input signals into output signals, and the translation is carried out using a processing chain that is adaptable by modifying values of internal processing parameters, wherein the training of the central AI module takes place by modifying the internal processing parameters based on further internal processing parameters of further AI modules, the further AI modules being in a plurality of vehicles and translating input signals into output signals in each case, and the translations taking place using processing chains that are able to be adapted by modifying values of further internal processing parameters, the further AI modules having been trained using input signals that are based on environment data acquired with using environment sensor systems installed in the vehicles.

First claim

What is claimed is:

1. A method for training a central artificial intelligence module (“AI module”), situated on a server, for highly or fully automated operation of a vehicle, comprising:
executing at least one learning step for training a first AI module situated in a first vehicle or a HIL (hardware in the loop) simulation of the first vehicle, based on input signals which are based on environment data acquired by an environment sensor system installed in the first vehicle or being based on recorded environment data of a vehicle, and first internal processing parameters of the first AI module are adapted during the training, on the basis of which input signals entered into the first AI module are translated into output signals;
executing at least one learning step for training a second AI module situated in a second vehicle or a HIL simulation of a second vehicle, based on input signals which are based on environment data acquired by an environment sensor system installed in the second vehicle or being based on recorded environment data of a vehicle, and second internal processing parameters of the second AI module are adapted during the training, based on which input signals entered into the second AI module are translated into output signals;
transmitting the adapted first internal processing parameters to the server;
transmitting the adapted second internal processing parameters to the server; and
training the central AI module based on the transmitted adapted first and second internal processing parameters, wherein only the adapted internal processing parameters of the at least one selected vehicle are utilized for training the central AI module by excluding the environment data acquired by the environment sensor system installed in the first vehicle, the environment data acquired by the environment sensor system installed in the second vehicle, and the recorded environment data from being transmitted to the central AI module.

2. The method as recited in claim 1 , wherein the central AI module translates input signals into output signals based on adaptable central internal processing parameters, and in the step of training the central AI module, the central internal processing parameters are adapted based on the adapted first and second internal processing parameters, wherein no input signals of the first or second vehicle or the first or second HIL simulation are used for adapting the central internal processing parameters.

3. The method as recited in claim 1 , wherein individual driving functions and/or driving maneuvers are learned with the aid of the first and/or second AI module.

4. The method as recited in claim 1 , wherein identical internal output processing parameters are selected for the first and the second AI module, the internal output processing parameters corresponding to the internal processing parameters prior to a first learning process of the AI modules, the first and the second internal output processing parameters having identical output values.

5. The method as recited in claim 1 , wherein the server receives adapted internal processing parameters from further vehicles having AI modules, and the method includes the additional step of selecting at least one of the further vehicles.

6. The method as recited in claim 1 , further comprising:
transmitting the central AI module to at least one vehicle.

7. The method as recited in claim 6 , further comprising the highly or fully automated operation of the vehicle using the transmitted central AI module.

8. The method as recited in claim 1 , wherein environment data acquired with the aid of the environment sensor system of the first and/or the second vehicle are buffer-stored in the first and/or the second vehicle and the first and/or the second AI module is/are trained based on the buffer-stored environment data, the data being buffer-stored for more than ten seconds.

9. The method as recited in claim 8 , wherein the data is buffer-stored for more than 10 minutes.

10. The method as recited in claim 9 , wherein the data is buffer-stored for more than 24 hours.

11. A method for training a central artificial intelligence module (“AI module”) situated on a server, for highly or fully automated operation of a vehicle, comprising:
receiving values of first adapted internal processing parameters of a first AI module from a first vehicle and/or a first HIL simulation, whereby input signals entered into the first AI module are translated into output signals based on the first internal processing parameters, and whereby at least one learning step in which the first internal processing parameters were adapted was already performed by the first AI module, and the adaptation of the first internal processing parameters has been carried out based on environment data acquired with the aid of an environment sensor system installed in the first vehicle or based on recorded environment data of a vehicle;
receiving values of second adapted internal processing parameters of a second AI module from a second vehicle and/or a second HIL simulation, whereby input signals entered into the second AI module are translated into output signals based on the second internal processing parameters, and whereby at least one learning step in which the second internal processing parameters were adapted was already performed by the second AI module, and the adaptation of the second internal processing parameters has been carried out based on environment data acquired with the aid of an environment sensor system installed in the second vehicle or based on recorded environment data of a vehicle; and
training the central AI module based on the received adapted first and second internal processing parameters, wherein only the adapted internal processing parameters of the at least one selected vehicle are utilized for training the central AI module by excluding the environment data acquired by the environment sensor system installed in the first vehicle, the environment data acquired by the environment sensor system installed in the second vehicle, and the recorded environment data from being transmitted to the central AI module.

12. A central artificial intelligences module (“AI module”) for the highly or fully automated operation of a vehicle, the central AI module being configured to translate input signals into output signals, and the translation takes place on the basis of internal processing parameters, wherein the central AI module is trained by:
receiving values of first adapted internal processing parameters of a first AI module from a first vehicle and/or a first HIL simulation, whereby input signals entered into the first AI module are translated into output signals based on the first internal processing parameters, and whereby at least one learning step in which the first internal processing parameters were adapted was already performed by the first AI module, and the adaptation of the first internal processing parameters has been carried out based on environment data acquired with the aid of an environment sensor system installed in the first vehicle or based on recorded environment data of a vehicle;
receiving values of second adapted internal processing parameters of a second AI module from a second vehicle and/or a second HIL simulation, whereby input signals entered into the second AI module are translated into output signals based on the second internal processing parameters, and whereby at least one learning step in which the second internal processing parameters were adapted was already performed by the second AI module, and the adaptation of the second internal processing parameters has been carried out based on environment data acquire

Assignees

• Bosch Gmbh Robert

Inventors

• Ewert Marlon Ramon

Classifications

• G06N3/09 Primary

  Supervised learning · CPC title

• G06N3/098

  Distributed learning, e.g. federated learning · CPC title

• G06N3/092

  Reinforcement learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• H04L67/10

  in which an application is distributed across nodes in the network (software deployment G06F8/60; multiprogramming arrangements G06F9/46) · CPC title