Machine learning models for adjusting communication parameters

1. A method performed by one or more computers, the method comprising:

obtaining, by the one or more computers, data for each terminal in a set of multiple satellite terminals, the data for each terminal indicating (i) a current satellite beam used by the terminal, (ii) an efficiency measure for communication of the terminal using the current satellite beam, and (iii) a geographic location of the terminal;

training, by the one or more computers, a machine learning model based on the obtained data to receive an indication of a geographic location and predict a satellite beam capable of providing at least a minimum level of efficiency for communication at the geographic location;

after training the machine learning model, generating, using the machine learning model, an indication of a predicted satellite beam for a particular geographic location, the predicted satellite beam being predicted by the trained machine learning model to provide at least the minimum level of efficiency for communication by a terminal at the particular geographic location; and

determining, by the one or more computers, whether to change the current satellite beam for a terminal at the particular geographic location based on the predicted satellite beam.

2. The method of claim 1 , wherein the machine learning model is trained to select, from among multiple satellite beams, a satellite beam providing a highest efficiency of communication for the geographic location indicated to the machine learning model.

3. The method of claim 1 , wherein the efficiency measure for a terminal is based on a modulation or coding used by the terminal.

4. The method of claim 1 , wherein the efficiency measure for a terminal indicates a number of bits per symbol transferred using the current satellite beam assigned to the terminal.

5. The method of claim 1 , wherein the machine learning model is trained to predict a satellite beam for a terminal from among multiple beams of a single satellite.

6. The method of claim 1 , wherein the machine learning model is trained to predict a satellite beam for a terminal from among beams of multiple satellites.

7. The method of claim 1 , wherein determining whether to change the current satellite beam for a terminal at the particular geographic location comprises determining to assign the predicted satellite beam to the terminal at the particular geographic location; and

wherein the method includes initiating a change of satellite beam for the terminal at the particular geographic location from a current satellite beam to the predicted satellite beam.

8. The method of claim 1 , wherein training the machine learning model comprises:

identifying a subset of the multiple satellite terminals that have an efficiency measure that satisfies a threshold; and

using the data for the identified subset of satellite terminals as training examples for training the machine learning model;

wherein the method comprises, after training the machine learning model, using the trained machine learning model to determine a predicted satellite beam for each of the satellite terminals that have efficiency measures that do not satisfy the threshold.

9. The method of claim 8 , further comprising, selecting, from among the satellite terminals that have efficiency measures that do not satisfy the threshold, a set of candidate terminals for satellite beam change for which the predicted satellite beams for the terminals are different from the current satellite beams for the terminals;

evaluating the predicted satellite beams for the terminals based on locations of the candidate terminals; and

selecting, from among the candidate terminals, a final subset to be re-assigned to their respective predicted satellite beams.

10. The method of claim 9 , further comprising generating map data indicating the locations of the candidate terminals or the terminals in the final subset; and

providing the map data for display.

11. The method of claim 1 , wherein the machine learning model is a first machine learning model;

wherein the method further comprises training a second machine learning model based on the obtained data for the multiple satellite terminals, wherein the first machine learning model and the second machine learning model operates using a different predictive algorithm; and

wherein the indication of a predicted satellite beam for a particular location is generated using output of the first machine learning model and the second machine learning model.

12. The method of claim 1 , wherein the machine learning model comprises a neural network, a maximum entropy classifier, a decision tree, an XG boost tree, a random forest classifier, a support vector machine, or a logistic regression model.

13. The method of claim 1 , wherein the efficiency measure for a terminal is based on a compression ratio for the terminal or an end-to-end traffic response time for the terminal.

14. The method of claim 1 , wherein training the machine learning model comprises training the machine learning model to predict a satellite beam for a terminal based on receiving only an indication of the geographic location of the terminal as input;

wherein generating the indication of the predicted satellite beam for the particular location comprises:

providing data indicating the particular location as input to the trained machine learning model; and

receiving, as output of the trained machine learning model, data indicating the predicted satellite beam for the particular location.

15. The method of claim 1 , wherein determining whether to change the satellite beam for a terminal at the particular location comprises:

comparing an identifier for the predicted satellite beam with an identifier for the current satellite beam used by the terminal;

determining, based on the comparison, that the predicted satellite beam is different from the current satellite beam used by the terminal;

determining that the location of the terminal is within a threshold distance of a center of the predicted satellite beam; and

determining to change the satellite beam assigned to the terminal based on determining that the predicted satellite beam is different from the current satellite beam used by the terminal and determining that the location of the terminal is within the threshold distance of the center of the predicted satellite beam.

16. The method of claim 1 , further comprising training multiple different machine learning models for different sets of satellite beams, wherein the machine learning model for a set of satellite beams is trained based on data describing terminals currently assigned to use one of the satellite beams in the set.

17. A system comprising:

one or more computers; and

one or more computer-readable media storing instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising:

obtaining, by the one or more computers, data for each terminal in a set of multiple satellite terminals, the data for each terminal indicating (i) a current satellite beam used by the terminal, (ii) an efficiency measure for communication of the terminal using the current satellite beam, and (iii) a geographic location of the terminal;

training, by the one or more computers, a machine learning model based on the obtained data to receive an indication of a geographic location and predict a satellite beam capable of providing at least a minimum level of efficiency for communication at the geographic location;

after training the machine learning model, generating, using the machine learning model, an indication of a predicted satellite beam for a particular geographic location, the predicted satellite beam being predicted by the trained machine learning model to provide at least the minimum level of efficiency for communication by a terminal at the particular geographic location; and

determining, by the one or more computers, whether to change the current satellite beam for a terminal at the particular geographic location based on the predicted satellite beam.

18. The system of claim 17 , wherein the machine learning model is trained to select, from among multiple satellite beams, a satellite beam providing a highest efficiency of communication for the geographic location indicated to the machine learning model.

19. The method of claim 1 , wherein the efficiency measure for a terminal indicates a level of efficiency of communication that the terminal achieved using the current satellite beam; and

wherein training the machine learning model is performed using the obtained data as training examples indicating levels of efficiency achieved for different beams at different geographic locations, the machine learning model being trained.

20. One or more non-transitory computer-readable media storing instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising:

obtaining, by the one or more computers, data for each terminal in a set of multiple satellite terminals, the data for each terminal indicating (i) a current satellite beam used by the terminal, (ii) an efficiency measure for communication of the terminal using the current satellite beam, and (iii) a geographic location of the terminal;

training, by the one or more computers, a machine learning model based on the obtained data to receive an indication of a geographic location and predict a satellite beam capable of providing at least a minimum level of efficiency for communication at the geographic location;

after training the machine learning model, generating, using the machine learning model, an indication of a predicted satellite beam for a particular geographic location, the predicted satellite beam being predicted by the trained machine learning model to provide at least the minimum level of efficiency for communication by a terminal at the particular geographic location; and

determining, by the one or more computers, whether to change the current satellite beam for a terminal at the particular geographic location based on the predicted satellite beam.