Methods and systems for selecting machine learning models to predict distributed computing resources

Patent metadata

Field	Value
Publication number	US-2021181739-A1
Application number	US-201916710590-A
Country	US
Kind code	A1
Filing date	Dec 11, 2019
Priority date	Dec 11, 2019
Publication date	Jun 17, 2021
Grant date	—

Abstract

A method includes receiving a request from a vehicle to perform a computing task, selecting a machine learning model from among a plurality of machine learning models based at least in part on the request, and predicting an amount of computing resources needed to perform the computing task using the selected machine learning model.

First claim

What is claimed is:

1 . A method comprising:
receiving a request from a vehicle to perform a computing task; selecting a machine learning model from among a plurality of machine learning models based at least in part on the request; and predicting an amount of computing resources needed to perform the computing task using the selected machine learning model.

2 . The method of claim 1 , further comprising allocating the amount of computing resources equal to the amount of computing resources predicted to be needed to perform the computing task.

3 . The method of claim 1 , further comprising:
receiving, from the vehicle, information about a driving state of the vehicle; and selecting the machine learning model based at least in part on the driving state of the vehicle.

4 . The method of claim 3 , wherein the information about the driving state of the vehicle comprises a speed of the vehicle.

5 . The method of claim 1 , further comprising:
receiving, from the vehicle, information related to the computing task; and selecting the machine learning model based at least in part on the information related to the computing task; wherein the information related to the computing task comprises one or more of: a size of the computing task; a deadline of the computing task; an application type of the computing task; a frequency of the computing task; and whether the computing task is deadline sensitive.

6 . The method of claim 1 , further comprising:
determining whether a current allocation of resources is sufficient for performing the computing task; in response to the current allocation of resources being not sufficient for performing the computing task, allocating additional resources for performing the computing task; and in response to the current allocation of resources being sufficient for performing the computing task, reallocating currently allocated resources that are not needed for performing the computing task to another resource provider in a second region.

7 . The method of claim 5 , wherein the frequency of the computing task comprises a time based generation frequency.

8 . The method of claim 5 , wherein the frequency of the computing task comprises a mileage based generation frequency.

9 . The method of claim 5 , wherein the frequency of the computing task comprises a velocity based generation frequency.

10 . The method of claim 5 , wherein the frequency of the computing task comprises a non-periodic generation frequency.

11 . The method of claim 1 , further comprising:
performing the computing task; determining an amount of resources consumed while performing the computing task; and updating training data for the selected machine learning model based on the amount of resources consumed while performing the computing task.

12 . An edge server comprising:
one or more processors; one or more memory modules; and machine readable instructions stored in the one or more memory modules that, when executed by the one or more processors, cause the edge server to:
receive a request from a vehicle to perform a computing task;
receive a driving status of the vehicle;
select a machine learning model from among a plurality of machine learning models based at least in part on the request and the driving status; and
based on the selected machine leaning model, predict an amount of computing resources needed to perform the computing task.

13 . The edge server of claim 12 , wherein the machine readable instructions stored in the one or more memory modules, when executed by the one or more processors, cause the edge server to:
tune one or more hyperparameters of the selected machine learning model based at least in part on the request or the driving status.

14 . The edge server of claim 12 , wherein the machine readable instructions stored in the one or more memory modules, when executed by the one or more processors, cause the edge server to:
receive a task generation frequency related to the computing task from the vehicle; and select the machine learning model based at least in part on the received task generation frequency.

15 . The edge server of claim 12 , wherein the machine readable instructions stored in the one or more memory modules, when executed by the one or more processors, cause the edge server to:
receive data relating to an amount of resources consumed while performing the computing task; and update training data associated with the selected machine learning model based on the received data relating to the amount of resources consumed while performing the computing task.

16 . The edge server of claim 15 , wherein the machine readable instructions stored in the one or more memory modules, when executed by the one or more processors, cause the edge server to:
preprocess the data relating to the amount of resources consumed while performing the computing task before updating the training data associated with the selected machine learning model.

17 . A system comprising:
an edge server; a cloud server communicatively coupled to the edge server; and a resource manager communicatively coupled to the cloud server, wherein the edge server is configured to:
receive, from a vehicle, a request to perform a computing task and a driving status of the vehicle;
select a machine learning model from among a plurality of machine learning models based at least in part on the request and the driving status; and
predict an amount of computing resources needed to perform the computing task based on the selected machine learning model;
wherein the resource manager is configured to allocate an amount of computing resources for performing the computing task to the cloud server or the edge server; and wherein the cloud server is configured to use the allocated computing resources to perform the computing task.

18 . The system of claim 17 , wherein the edge server is configured to tune one or more hyperparameters of the selected machine learning model based at least in part on the request or the driving status.

19 . The system of claim 17 , wherein:
the cloud server is configured to monitor an amount of resources consumed while performing the computing task and transmit this information to the edge server; and the edge server is configured to update training data associated with the selected machine learning model based on the amount of resources consumed while performing the computing task.

20 . The system of claim 19 , wherein the cloud server is configured to preprocess data related to the amount of resources consumed while performing the computing task before updating the training data associated with the selected machine learning model.

Assignees

• Toyota Eng & Mfg North America

Inventors

• Chen Dawei
• Liu Yin-Chen
• Kim Baekgyu

Classifications

• H04L67/1012 Primary

  based on compliance of requirements or conditions with available server resources · CPC title

• G06F18/285

  Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system · CPC title

• G06F18/2413

  based on distances to training or reference patterns · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N3/048

  Activation functions · CPC title