Detecting radio coverage problems

1. A method comprising:

feeding, by data processing hardware, a predicted location of a base station into a neural network, the neural network configured to output an estimated characteristic of the base station and to output a confidence indicator for the estimated characteristic of the base station, the confidence indicator representing an uncertainty of the estimated characteristic;

determining, by the data processing hardware, the estimated characteristic of the base station by feeding a coverage map into the neural network, the coverage map indicating a signal characteristic of the emitted signal about the base station, the estimated characteristic of the base station comprising an estimated location of the base station;

determining, by the data processing hardware, a prediction error based on (i) a deviation between the predicted location of the base station and the estimated location of the base station and (ii) the confidence indicator associated with the estimated characteristic output by the neural network; and

communicating, by the data processing hardware, the prediction error to an issue queue at a network entity administrator.

2. The method of claim 1 , wherein the coverage map for the base station is based on observations, each observation comprising:

a signal measurement of a signal emitted from the base station to a corresponding user equipment (UE) device; and

a location of the corresponding UE device at a time the corresponding UE device either measures the signal measurement or transmits the observation.

3. The method of claim 2 , wherein the coverage map for the base station is generated by:

dividing a coverage area about the base station into pixels, each pixel corresponding to a geographical portion of the coverage area;

for each observation:

identifying the pixel having the corresponding geographical portion of the coverage area that contains the location of the observation; and

associating the observation with the identified pixel; and

for each pixel, averaging the signal measurements of any observations associated with a respective pixel.

4. The method of claim 2 , wherein determining the estimated characteristic of the base station further comprises feeding the coverage map and an observation map into the neural network, the observation map generated by, for each pixel of the coverage map, generating a metric that monotonically expresses a number of any observations associated with the respective pixel.

5. The method of claim 4 , wherein the metric comprises at least one of:

a number of any observations associated with the respective pixel;

a log of the number of any observations associated with the respective pixel; or

a monotonic function of the number of any observations associated with the respective pixel.

6. The method of claim 5 , wherein generating the metric comprises:

determining a number of any observations associated with the respective pixel;

when the number is greater than zero, assigning a value of the respective pixel to one; and

when the number equals zero, assigning the value of the respective pixel to zero.

7. The method of claim 2 , wherein determining the estimated characteristic of the base station further comprises feeding the coverage map and a location uncertainty map into the neural network, wherein each observation comprises a location uncertainty measurement, and the location uncertainty map is generated based on the location uncertainty measurements.

8. The method of claim 1 , further comprising feeding a terrain map of a geographical area about the base station into the neural network, the terrain map describing at least one of:

a terrain altitude of the geographical area; or

a presence and/or height of objects extending above a ground surface of the geographical area.

9. The method of claim 1 , further comprising feeding, by the data processing hardware, side information into the neural network, the side information comprising at least one of a frequency of operation of the base station, a height of an antenna of the base station, an antenna beam width, or an antenna tilt angle.

10. The method of claim 1 , wherein the estimated characteristic of the base station further comprises an estimated pointing direction of the base station or an antenna azimuth of the base station.

11. A system comprising:

data processing hardware; and

memory hardware in communication with the data processing hardware, the memory hardware storing instructions that when executed on the data processing hardware cause the data processing hardware to perform operations comprising:

feeding a predicted location of a base station into a neural network, the neural network configured to output an estimated characteristic of the base station and to output a confidence indicator for the estimated characteristic of the base station, the confidence indicator representing an uncertainty of the estimated characteristic;

determining the estimated characteristic of the base station by feeding a coverage map into the neural network, the coverage map indicating a signal characteristic of the emitted signal about the base station, the estimated characteristic of the base station comprising an estimated location of the base station;

determining a prediction error based on (i) a deviation between the predicted location of the base station and the estimated location of the base station and (ii) the confidence indicator associated with the estimated characteristic output by the neural network; and

communicating the prediction error to an issue queue at a network entity administrator.

12. The system of claim 11 , wherein the coverage map for the base station is based on observations, each observation comprising:

a signal measurement of a signal emitted from the base station to a corresponding user equipment (UE) device; and

a location of the corresponding UE device at a time the corresponding UE device either measures the signal measurement or transmits the observation.

13. The system of claim 12 , wherein the coverage map for the base station is generated by:

dividing a coverage area about the base station into pixels, each pixel corresponding to a geographical portion of the coverage area;

for each observation:

identifying the pixel having the corresponding geographical portion of the coverage area that contains the location of the observation; and

associating the observation with the identified pixel; and

for each pixel, averaging the signal measurements of any observations associated with a respective pixel.

14. The system of claim 12 , wherein determining the estimated characteristic of the base station further comprises feeding the coverage map and an observation map into the neural network, the observation map generated by, for each pixel of the coverage map, generating a metric that monotonically expresses a number of any observations associated with the respective pixel.

15. The system of claim 14 , wherein the metric comprises at least one of:

a number of any observations associated with the respective pixel;

a log of the number of any observations associated with the respective pixel; or

a monotonic function of the number of any observations associated with the respective pixel.

16. The system of claim 15 , wherein generating the metric comprises:

determining a number of any observations associated with the respective pixel;

when the number is greater than zero, assigning a value of the respective pixel to one; and

when the number equals zero, assigning the value of the respective pixel to zero.

17. The system of claim 12 , wherein determining the estimated characteristic of the base station further comprises feeding the coverage map and a location uncertainty map into the neural network, wherein each observation comprises a location uncertainty measurement, and the location uncertainty map is generated based on the location uncertainty measurements.

18. The system of claim 11 , further comprising feeding a terrain map of a geographical area about the base station into the neural network, the terrain map describing at least one of:

a terrain altitude of the geographical area; or

a presence and/or height of objects extending above a ground surface of the geographical area.

19. The system of claim 11 , further comprising feeding, by the data processing hardware, side information into the neural network, the side information comprising at least one of a frequency of operation of the base station, a height of an antenna of the base station, an antenna beam width, or an antenna tilt angle.

20. The system of claim 11 , wherein the estimated characteristic of the base station further comprises an estimated pointing direction of the base station or an antenna azimuth of the base station.