Utilizing invariant shadow fading data for training a machine learning model
A device may receive real mobile radio data identifying measurements of radio transmissions of base stations and user devices of a mobile radio environment in a geographical area, and may receive network topology data associated with the geographical area. The device may utilize, based on the network topology data, a machine learning feature extraction approach to generate a representation of invariant aspects of spatiotemporal predictable components of the real mobile radio data, and may generate, based on the representation of invariant aspects, stochastic data that includes a probability that a radio signal will be obstructed. The device may utilize the stochastic data to identify a realistic discoverable spatiotemporal signature, and may train or evaluate a system to manage performance of a mobile radio network based on the realistic discoverable spatiotemporal signature.
1. A method, comprising:
receiving, by a device, real mobile radio data identifying measurements of radio transmissions of base stations and user devices of a mobile radio environment in a geographical area;
receiving, by the device, network topology data associated with the geographical area;
utilizing, by the device and based on the network topology data, a machine learning feature extraction approach to generate a representation of invariant aspects of spatiotemporal predictable components of the real mobile radio data;
generating, by the device and based on the representation of invariant aspects, stochastic data that includes a probability that a radio signal will experience fading;
utilizing, by the device, the stochastic data to identify a realistic discoverable spatiotemporal signature;
training or evaluating, by the device, a system to manage performance of a mobile radio network based on the realistic discoverable spatiotemporal signature;
generating a validation dataset for the system based on the realistic discoverable spatiotemporal signature; and
validating the system with the validation dataset.
2. The method of claim 1 , further comprising:
utilizing a statistical analysis approach to identify a shadow decorrelation distance and a shadow standard deviation of the real mobile radio data,
wherein utilizing the machine learning feature extraction approach to generate the representation of invariant aspects includes:
generating the representation of invariant aspects based on the shadow decorrelation distance and the shadow standard deviation of the real mobile radio data.
3. The method of claim 1 , further comprising:
utilizing a statistical analysis approach to identify a shadow decorrelation distance and a shadow standard deviation relating to another geographical area,
wherein utilizing the machine learning feature extraction approach to generate the representation of invariant aspects includes:
generating another representation of invariant aspects based on the shadow decorrelation distance and the shadow standard deviation relating to the other geographical area,
wherein the other representation of the invariant aspects of spatiotemporal predictable components of the real mobile radio data of the geographical area is used as an input to the machine learning feature extraction approach.
4. The method of claim 1 , further comprising:
generating a test dataset for the system based on the realistic discoverable spatiotemporal signature; and
testing the system with the test dataset.
5. The method of claim 1 , further comprising:
training a machine learning model with the realistic discoverable spatiotemporal signature to generate a trained machine learning model; and
causing the trained machine learning model to be implemented by an element collocated with a network device associated with the mobile radio environment,
wherein the element is configured to utilize the trained machine learning model to identify shadow fading features in a coverage area of the network device.
6. The method of claim 1 , wherein the real mobile radio data includes a composite waveform with a slow fading component caused by permanent obstructions, a slow fading component caused by temporal obstructions, and a fast fading component.
7. The method of claim 1 , wherein the network topology data includes obstruction data identifying permanent obstructions in the geographical area and temporal obstructions in the geographical area.
8. A device, comprising:
one or more memories; and
one or more processors, coupled to the one or more memories, configured to:
receive real mobile radio data identifying measurements of radio transmissions of base stations and user devices of a mobile radio environment in a geographical area;
receive network topology data associated with the geographical area;
utilize, based on the network topology data, a machine learning feature extraction approach to generate a representation of invariant aspects of spatiotemporal predictable components of the real mobile radio data;
generate, based on the representation of invariant aspects, stochastic data that includes a probability that a radio signal will experience fading;
utilize the stochastic data to identify a realistic discoverable spatiotemporal signature;
train or evaluate a system to manage performance of a mobile radio network based on the realistic discoverable spatiotemporal signature;
generate a validation dataset for the system based on the realistic discoverable spatiotemporal signature; and
validate the system with the validation dataset.
9. The device of claim 8 , wherein the one or more processors are further configured to:
utilize a statistical analysis approach to identify a shadow decorrelation distance and a shadow standard deviation relating to another geographical area; and
generate another representation of invariant aspects based on the shadow decorrelation distance and the shadow standard deviation relating to the other geographical area,
wherein the other representation of the invariant aspects of spatiotemporal predictable components of the real mobile radio data of the geographical area is used as an input to the machine learning feature extraction approach.
10. The device of claim 8 , wherein the one or more processors are further configured to:
generate a test dataset for the system based on the realistic discoverable spatiotemporal signature; and
test the system with the test dataset.
11. The device of claim 8 , wherein the one or more processors are further configured to:
train a machine learning model with the realistic discoverable spatiotemporal signature to generate a trained machine learning model; and
cause the trained machine learning model to be implemented by an element collocated with a network device associated with the mobile radio environment,
wherein the element is configured to utilize the trained machine learning model to identify shadow fading features in a coverage area of the network device.
12. The device of claim 8 , wherein the real mobile radio data includes a composite waveform with a slow fading component caused by permanent obstructions, a slow fading component caused by temporal obstructions, and a fast fading component.
13. The device of claim 8 , wherein the network topology data includes obstruction data identifying permanent obstructions in the geographical area and temporal obstructions in the geographical area.
14. The device of claim 8 , wherein the one or more processors are further configured to:
utilize a statistical analysis approach to identify a shadow decorrelation distance and a shadow standard deviation of the real mobile radio data,
wherein the one or more processors, to generate the representation of invariant aspects, are configured to:
generate the representation of invariant aspects based on the shadow decorrelation distance and the shadow standard deviation of the real mobile radio data.
15. A non-transitory computer-readable medium storing a set of instructions, the set of instructions comprising:
one or more instructions that, when executed by one or more processors of a device, cause the device to:
receive real mobile radio data identifying measurements of radio transmissions of base stations and user devices of a mobile radio environment in a geographical area;
receive network topology data associated with the geographical area;
utilize, based on the network topology data, a machine learning feature extraction approach to generate a representation of invariant aspects of spatiotemporal predictable components of the real mobile radio data;
generate, based on the representation of invariant aspects, stochastic data that includes a probability that a radio signal will experience fading;
utilize the stochastic data to identify a realistic discoverable spatiotemporal signature;
train or evaluate a system to manage performance of a mobile radio network based on the realistic discoverable spatiotemporal signature;
generate a validation dataset for the system based on the realistic discoverable spatiotemporal signature; and
validate the system with the validation dataset.
16. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions further cause the device to:
utilize a statistical analysis approach to identify a shadow decorrelation distance and a shadow standard deviation relating to another geographical area; and
generate another representation of invariant aspects based on the shadow decorrelation distance and the shadow standard deviation relating to the other geographical area,
wherein the other representation of the invariant aspects of spatiotemporal predictable components of the real mobile radio data of the geographical area is used as an input to the machine learning feature extraction approach.
17. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions further cause the device to:
generate a test dataset for the system based on the realistic discoverable spatiotemporal signature; and
test the system with the test dataset.
18. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions further cause the device to:
train a machine learning model with the realistic discoverable spatiotemporal signature to generate a trained machine learning model; and
cause the trained machine learning model to be implemented by an element collocated with a network device associated with the mobile radio environment,
wherein the element is configured to utilize the trained machine learning model to identify shadow fading features in a coverage area of the network device.
19. The non-transitory computer-readable medium of claim 15 , wherein the real mobile radio data includes a composite waveform with a slow fading component caused by permanent obstructions, a slow fading component caused by temporal obstructions, and a fast fading component.
20. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions further cause the device to:
utilize a statistical analysis approach to identify a shadow decorrelation distance and a shadow standard deviation of the real mobile radio data,
wherein the one or more instructions, that cause the device to generate the representation of invariant aspects, cause the device to:
generate the representation of invariant aspects based on the shadow decorrelation distance and the shadow standard deviation of the real mobile radio data.