System and method for applying machine learning to mobile geolocation

1. A communication network monitoring device comprising a processor and a memory, wherein the memory comprises instructions that, when executed by the processor, cause the device to:

receive, from a base station (BS), radio predictors of a first user equipment (UE) for which a minimization of drive test (MDT) mode is activated, a UE location history of the first UE, a geolocation of the first UE, radio predictors of a second UE for which the MDT mode is not activated, a UE location history of the second UE, and cell physical parameters, wherein the radio predictors of the first UE comprise a timing advance (TA) of the first UE and a channel quality indication (CQI) of the first UE, and wherein the radio predictors of the second UE comprise a TA of the second UE and a CQI of the second UE;

train a machine learning (ML) model to determine an azimuth of the second UE at least based on the radio predictors of the first UE, the UE location history of the first UE, the geolocation of the first UE, and the cell physical parameters;

execute the ML model to determine the azimuth of the second UE; and

provide, to a downstream application, a geolocation of the second UE at least based on the azimuth of the second UE and the TA of the second UE.

2. The device of claim 1 , wherein the ML model that is executed to determine the azimuth of the second UE is trained using the TA of the first UE.

3. The device of claim 1 , the instructions, when executed by the processor, further cause the device to:

train the ML model using the TA of the first UE before training the ML model to determine the azimuth of the second UE; and

execute the ML model that was trained using the TA of the first UE to generate a machine learning timing advance (MLTA) of the first UE;

wherein training the ML model to determine the azimuth of the second UE comprises training the ML model using the MLTA of the first UE,

wherein executing the ML model to determine the azimuth of the second UE comprises executing the ML model that was trained using the MLTA of the first UE to determine the azimuth of the second UE.

4. The device of claim 3 , the instructions, when executed by the processor, further cause the device to:

execute the ML model that was trained using the TA of the first UE to generate a machine learning timing advance (MLTA) of the second UE;

provide, to a downstream application, a geolocation of the second UE at least based on the MLTA of the second UE.

5. The device of claim 1 , the instructions, when executed by the processor, further cause the device to:

train a second ML model using the TA of the first UE before training the ML model to determine the azimuth of the second UE, wherein the ML model uses a first algorithm and the second ML model uses a second algorithm different from the first algorithm; and

execute the second ML model that was trained using the TA of the first UE to generate a machine learning timing advance (MLTA) of the first UE;

wherein training the ML model to determine the azimuth of the second UE comprises training the ML model using the MLTA of the first UE.

6. The device of claim 1 , the instructions, when executed by the processor, further cause the device to:

determine an estimated distance between the BS and the second UE at least based on the TA of the second UE,

wherein providing, to the downstream application, the geolocation of the second UE at least based on the azimuth of the second UE and the TA of the second UE comprises providing, to the downstream application, the geolocation of the second UE at least based on the azimuth of the second UE and the estimated distance between the BS and the second UE.

7. The device of claim 1 , wherein providing, to the downstream application, the geolocation of the second UE at least based on the azimuth of the second UE and the TA of the second UE comprises:

converting the azimuth and an estimated distance between the BS and the second UE at least based on the TA of the second UE into a latitude coordinate and a longitude coordinate; and

providing the latitude coordinate and the longitude coordinate to the downstream application.

8. The method of claim 1 , wherein the ML model uses a plurality of decision trees, wherein the plurality of decision trees are used in accordance with the Random Forest algorithm.

9. The device of claim 1 , wherein the radio predictors of the first UE comprise a measurement report of the first UE, and wherein the radio predictors of the second UE comprise a measurement report of the second UE.

10. The device of claim 9 , wherein the measurement report of the first UE comprises a serving cell Reference Signal Received Power (RSRP) of the first UE, a serving cell Reference Signal Received Quality (RSRQ) of the first UE, a neighbor cell RSRP of the first UE, a neighbor cell RSRQ of the first UE, and MDT data, and wherein the measurement report of the second UE comprises a serving cell RSRP of the first UE, a serving cell RSRQ of the second UE, a neighbor cell RSRP of the second UE, and a neighbor cell RSRQ of the second UE.

11. The device of claim 10 , wherein the measurement report of the first UE comprises a serving cell synchronization signal/physical broadcast channel (PBCH) block (SSB) index of the first UE, and a neighbor cell SSB index of the first UE, and wherein the measurement report of the second UE comprises a serving cell SSB index of the second UE, and a neighbor cell SSB index of the second UE.

12. The device of claim 1 , wherein the cell physical parameters comprise an antenna latitude location, an antenna longitude location, an antenna horizontal beam width, an antenna vertical beam width, an antenna boresight direction, an antenna gain and an antenna signal transmit power.

13. A method to determine a geolocation, the method comprising:

receiving, by a processor, from a base station (BS), radio predictors of a first user equipment (UE) for which a minimization of drive test (MDT) mode is activated, a UE location history of the first UE, a geolocation of the first UE, radio predictors of a second UE for which the MDT mode is not activated, a UE location history of the second UE, and cell physical parameters, wherein the radio predictors of the first UE comprise a timing advance (TA) of the first UE and a channel quality indication (CQI) of the first UE, and wherein the radio predictors of the second UE comprise a TA of the second UE and a CQI of the second UE;

training, by the processor, a machine learning (ML) model to determine an azimuth of the second UE at least based on the radio predictors of the first UE, the UE location history of the first UE, the geolocation of the first UE, and the cell physical parameters;

executing, by the processor, the ML model to determine the azimuth of the second UE; and

providing, by the processor, to a downstream application, a geolocation of the second UE at least based on the azimuth of the second UE and the TA of the second UE.

14. The method of claim 13 , wherein the ML model that is executed to determine the azimuth of the second UE is trained using the TA of the first UE.

15. The method of claim 13 , further comprising:

training the ML model using the TA of the first UE before training the ML model to determine the azimuth of the second UE; and

executing the ML model that was trained using the TA of the first UE to generate a machine learning timing advance (MLTA) of the first UE;

wherein training the ML model to determine the azimuth of the second UE comprises training the ML model using the MLTA of the first UE,

wherein executing the ML model to determine the azimuth of the second UE comprises executing the ML model that was trained using the MLTA of the first UE to determine the azimuth of the second UE.

16. The method of claim 13 , wherein providing, to the downstream application, the geolocation of the second UE at least based on the azimuth of the second UE and the TA of the second UE comprises:

converting the azimuth and an estimated distance between the BS and the second UE at least based on the TA of the second UE into a latitude coordinate and a longitude coordinate; and

providing the latitude coordinate and the longitude coordinate to the downstream application.

17. The method of claim 13 , wherein the radio predictors of the first UE comprise a measurement report of the first UE, and wherein the radio predictors of the second UE comprise a measurement report of the second UE.

18. The method of claim 17 , wherein the measurement report of the first UE comprises a serving cell Reference Signal Received Power (RSRP) of the first UE, a serving cell Reference Signal Received Quality (RSRQ) of the first UE, a neighbor cell RSRP of the first UE, a neighbor cell RSRQ of the first UE, and MDT data, and wherein the measurement report of the second UE comprises a serving cell RSRP of the first UE, a serving cell RSRQ of the second UE, a neighbor cell RSRP of the second UE, and a neighbor cell RSRQ of the second UE.

19. A non-transitory computer-readable medium comprising instructions stored thereon that, when executed by a processor, cause the processor to:

receive, from a base station (BS), radio predictors of a first user equipment (UE) for which a minimization of drive test (MDT) mode is activated, a UE location history of the first UE, a geolocation of the first UE, radio predictors of a second UE for which the MDT mode is not activated, a UE location history of the second UE, and cell physical parameters, wherein the radio predictors of the first UE comprise a timing advance (TA) of the first UE and a channel quality indication (CQI) of the first UE, and wherein the radio predictors of the second UE comprise a TA of the second UE and a CQI of the second UE;

train a machine learning (ML) model to determine an azimuth of the second UE at least based on the radio predictors of the first UE, the UE location history of the first UE, the geolocation of the first UE, and the cell physical parameters;

execute the ML model to determine the azimuth of the second UE; and

provide, to a downstream application, a geolocation of the second UE at least based on the azimuth of the second UE and the TA of the second UE.

20. The medium of claim 19 , wherein the ML model that is executed to determine the azimuth of the second UE is trained using the TA of the first UE.

21. The medium of claim 19 , the instructions, when executed by the processor, further cause the processor to:

train the ML model using the TA of the first UE before training the ML model to determine the azimuth of the second UE; and

execute the ML model that was trained using the TA of the first UE to generate a machine learning timing advance (MLTA) of the first UE;

wherein training the ML model to determine the azimuth of the second UE comprises training the ML model using the MLTA of the first UE,

wherein executing the ML model to determine the azimuth of the second UE comprises executing the ML model that was trained using the MLTA of the first UE to determine the azimuth of the second UE.

22. The medium of claim 19 , wherein providing, to the downstream application, the geolocation of the second UE at least based on the azimuth of the second UE and the TA of the second UE comprises:

converting the azimuth and an estimated distance between the BS and the second UE at least based on the TA of the second UE into a latitude coordinate and a longitude coordinate; and

providing the latitude coordinate and the longitude coordinate to the downstream application.

23. The medium of claim 19 , wherein the radio predictors of the first UE comprise a measurement report of the first UE, and wherein the radio predictors of the second UE comprise a measurement report of the second UE.

24. The medium of claim 23 , wherein the measurement report of the first UE comprises a serving cell Reference Signal Received Power (RSRP) of the first UE, a serving cell Reference Signal Received Quality (RSRQ) of the first UE, a neighbor cell RSRP of the first UE, a neighbor cell RSRQ of the first UE, and MDT data, and wherein the measurement report of the second UE comprises a serving cell RSRP of the first UE, a serving cell RSRQ of the second UE, a neighbor cell RSRP of the second UE, and a neighbor cell RSRQ of the second UE.