Bandwidth allocation using machine learning

Patent metadata

Field	Value
Publication number	US-11503615-B2
Application number	US-201916732252-A
Country	US
Kind code	B2
Filing date	Dec 31, 2019
Priority date	Dec 31, 2019
Publication date	Nov 15, 2022
Grant date	Nov 15, 2022

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for bandwidth allocation using machine learning. In some implementations, a request for bandwidth in a communications system is received. Data indicative of a measure of bandwidth requested and a status of the communication system are provided as input to a machine learning model. One or more outputs from the machine learning model indicate an amount of bandwidth to allocate to the terminal, and bandwidth is allocated to the terminal based on the one or more outputs from the machine learning model.

First claim

What is claimed is:

1. A method comprising:
receiving a request for bandwidth in a communications system, the request being associated with a terminal;
accessing data indicating a status of the communication system, wherein the status indicates at least one of an aggregate demand for the communication system, a throughput for the communication system, or an available bandwidth for the communication system;
in response to receiving the request, providing, as input to a machine learning model, data indicative of (i) a measure of bandwidth requested for the terminal and (ii) a measure indicating at least one of the aggregate demand for the communication system, the throughput for the communication system, or the available bandwidth for the communication system, wherein the machine learning model has been trained to predict an allocation of bandwidth based on data indicative of an amount of data to be transferred;
receiving one or more outputs from the machine learning model that indicate an amount of bandwidth to allocate to the terminal; and
allocating bandwidth to the terminal based on the one or more outputs from the machine learning model.

2. The method of claim 1 , wherein the communication system comprises a satellite communication system.

3. The method of claim 1 , wherein the machine learning model has been trained using a set of training examples that each include a training target set using output of an allocation algorithm, wherein the allocation algorithm allocates bandwidth for a communication frame among terminals in a sequential manner such that bandwidth allocations are made in dependence on allocations made for other terminals for the communication frame; and
wherein the machine learning model is configured to predict a bandwidth allocations for terminals in a manner that is not dependent on allocations for other terminals.

4. The method of claim 1 , wherein the machine learning model is configured to predict allocation based on input data for a first set of parameters; and
wherein the machine learning model has been trained using training examples having training targets set using output generated by an allocation algorithm that receives input data for a second set of parameters, wherein the second set of parameters includes at least one parameter that is not included in the first set of parameters.

5. The method of claim 1 , wherein the communication system is a satellite communication system, and the request is from the terminal;
wherein the measure of bandwidth requested for the terminal is based on an amount of data transfer backlog for the terminal that is reported by the terminal to the satellite communication system, wherein the input to the machine learning model includes values indicative of data transfer backlog for the terminal for each of multiple priority levels; and
wherein the one or more outputs from the machine learning model comprise multiple outputs indicating amounts of bandwidth to allocate to the terminal for each of multiple priority levels.

6. The method of claim 1 , wherein the one or more outputs from the machine learning model indicate a number of slots in a time division multiple access (TDMA) communication frame to allocate to the terminal; and
wherein allocating bandwidth to the terminal comprises allocating a number of slots determined based on the one or more outputs to the terminal in the TDMA communication frame.

7. The method of claim 1 , wherein the machine learning model comprises at least one of a neural network, a classifier, a decision tree, a support vector machine, a regression model, a nearest neighbor method such as K-means or K-nearest neighbor, a dimensionality reduction algorithm, or a boosting algorithm.

8. The method of claim 1 , comprising:
determining a number of terminals connected using the communication system or a processor utilization for a device of the communication system; and
determining that the number of terminals or the processor utilization exceeds a threshold;
wherein allocating bandwidth to the terminal based on the one or more outputs from the machine learning model is performed at least in part based on determining that the number of terminals or the processor utilization exceeds a threshold.

9. The method of claim 1 , wherein the machine learning model is provided, in the input to a machine learning model used to determine the one or more outputs for the terminal, at least one of a priority of data to be transferred, a type of data to be transferred, a bandwidth limit associated with the terminal, a terminal identifier for the terminal, a quality of service level, or an error correction rate.

10. The method of claim 1 , comprising running multiple instances of the model in parallel to concurrently generate allocation predictions for each of multiple terminals receiving service from the communication system during a communication frame.

11. The method of claim 1 , wherein the input to the machine learning model comprises a measure indicating current demand of the communication system or current throughput of the communication system.

12. The method of claim 1 , wherein the input to the machine learning model comprises a measure indicating prior demand of the communication system or prior throughput of the communication system.

13. The method of claim 1 , wherein the input to the machine learning model comprises a measure indicating a data transfer capacity of the communication system or an available bandwidth of the communication system.

14. The method of claim 1 , wherein the request is a request for the terminal to transmit data in the communication system;
wherein the method includes receiving an inroute backlog report for the terminal indicating an amount of data queued for transmission by the terminal;
wherein the input to the machine learning model includes a value indicating the amount of data queued for transmission by the terminal; and
wherein the one or more outputs from the machine learning model indicate an amount of bandwidth to allocate for the terminal to transmit data in a communication frame.

15. The method of claim 1 , comprising, for a particular communication frame:
providing a set of input to the machine learning model for each of multiple terminals, wherein the sets of input respectively include (i) different values of requested bandwidth for different terminals and (ii) a same value indicating at least one of the aggregate demand for the communication system, the throughput for the communication system, or the available bandwidth;
receiving an output of the machine learning model for each of the multiple terminals, each of the outputs indicating a predicted bandwidth allocation determined for the corresponding terminal; and
assigning bandwidth in the particular communication frame to each of the multiple terminals, wherein the allocations for the multiple terminals are based on respective outputs of the model for the terminals.

16. The method of claim 15 , wherein assigning bandwidth in the particular communication frame comprises:
determining, based on the outputs of the machine learning model for the multiple terminals, a total amount of bandwidth to be allocated for the multiple terminals for a particular communication frame;
determining an amount of available bandwidth of the communication system for the particular communication frame;
scaling amounts of bandwidth that the outputs of the machine learning model indicate for multiple terminals based on (i) the total amount of bandwidth and (ii) the amount of available bandwidth; and
allocating, to the multiple termina

Assignees

• Hughes Network Systems Llc

Inventors

• Hu Bin
• Tang Yeqing
• Roy Rajarshi

Classifications

• H04W72/543 Primary

  based on requested quality, e.g. QoS · CPC title

• H04B7/18513

  Transmission in a satellite or space-based system · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• H04J3/1694

  Allocation of channels in TDM/TDMA networks, e.g. distributed multiplexers (Passive Optical Networks H04Q11/0062) · CPC title