Hybrid statistical and machine learning method and system for radio frequency (RF) spectrum propagation modeling and prediction
Aspects of the subject disclosure may include, for example, obtaining data regarding an environment associated with a cell site, identifying one or more coverage pivots from a plurality of coverage locations in the environment, utilizing one or more machine learning (ML) models to determine emission directions for the one or more coverage pivots based at least in part on locations of the one or more coverage pivots and the cell site in the environment, resulting in determined emission directions, and causing ray tracing simulation to be performed using the determined emission directions, wherein the ray tracing simulation enables first signal coverage estimates to be derived for the one or more coverage pivots and used to extrapolate second signal coverage estimates for a remainder of the plurality of coverage locations. Other embodiments are disclosed.
1 . A method, comprising:
obtaining, by a processing system including a processor, data regarding an environment associated with a cell site;
identifying, by the processing system, one or more coverage pivots from a plurality of coverage locations in the environment;
utilizing, by the processing system, one or more machine learning (ML) models to determine emission directions for the one or more coverage pivots based at least in part on locations of the one or more coverage pivots and the cell site in the environment, resulting in determined emission directions; and
causing, by the processing system, ray tracing simulation to be performed using the determined emission directions, wherein the ray tracing simulation enables first signal coverage estimates to be derived for the one or more coverage pivots and used to extrapolate second signal coverage estimates for a remainder of the plurality of coverage locations.
2 . The method of claim 1 , wherein the ray tracing simulation involves only the determined emission directions and not any other emission directions, thereby providing for reduced ray tracing as compared to a case where the other emission directions are also simulated.
3 . The method of claim 1 , wherein the one or more ML models comprise one or more deep learning models.
4 . The method of claim 3 , wherein the emission directions include azimuth angles and elevation angles, and wherein the one or more deep learning models include a first deep learning model trained for azimuth angles and a second deep learning model trained for elevation angles.
5 . The method of claim 3 , wherein the utilizing involves inputting, to linear layers of the one or more deep learning models, point cloud information relating to the environment, angle information, information regarding a location of the cell site, information regarding a coverage pivot, or a combination thereof.
6 . The method of claim 3 , wherein the utilizing involves utilizing the one or more deep learning models to determine, for each of the one or more coverage pivots, whether to emit or not in one or more angles of a plurality of angles.
7 . The method of claim 6 , wherein the plurality of angles comprises a quantized set of angles.
8 . The method of claim 3 , wherein the one or more deep learning models include a concatenation layer and a plurality of classification layers.
9 . The method of claim 1 , wherein the emission directions comprise photon emission directions.
10 . The method of claim 1 , wherein the one or more ML models are further configured to determine one or more photon emission budgets.
11 . The method of claim 1 , wherein the first signal coverage estimates, the second signal coverage estimates, or both comprise reference signal received power (RSRP) estimates.
12 . A device, comprising:
a processing system including a processor; and
a memory that stores executable instructions that, when executed by the processing system, facilitate performance of operations, the operations comprising:
receiving data regarding an environment associated with a cell tower;
selecting one or more coverage pivots from a plurality of coverage locations in the environment;
employing one or more deep learning models to determine photon emission directions for the one or more coverage pivots based at least in part on locations of the one or more coverage pivots and the cell tower, resulting in determined photon emission directions; and
causing a ray tracing simulator to simulate ray tracing only for the determined photon emission directions, wherein the ray tracing enables first signal coverage estimates to be derived for the one or more coverage pivots and used to extrapolate second signal coverage estimates for a remainder of the plurality of coverage locations.
13 . The device of claim 12 , wherein the photon emission directions include azimuth angles and elevation angles.
14 . The device of claim 12 , wherein the one or more deep learning models include a deep learning model trained for predicting whether emissions are to be made in an azimuth direction.
15 . The device of claim 12 , wherein the one or more deep learning models include a deep learning model trained for predicting whether emissions are to be made in an elevation direction.
16 . The device of claim 12 , wherein the employing involves inputting, to linear layers of the one or more deep learning models, point cloud information relating to the environment, angle information, information regarding a location of the cell tower, information regarding a coverage pivot, or a combination thereof.
17 . A non-transitory machine-readable medium, comprising executable instructions that, when executed by a processing system including a processor, facilitate performance of operations, the operations comprising:
receiving information regarding determined emission directions; and
performing ray tracing simulations based on the information, wherein the determined emission directions are output by one or more deep learning models that are trained to determine directions in which emissions are to be made, and wherein the one or more deep learning models are utilized to determine directions in which emissions are to be made for one or more coverage pivots selected from a plurality of coverage locations in an environment that includes a cell site, resulting in the determined emission directions.
18 . The non-transitory machine-readable medium of claim 17 , wherein the performing comprises performing the ray tracing simulations using only the determined emission directions and not any other emission directions, thereby providing for reduced ray tracing as compared to a case where the other emission directions are also simulated.
19 . The non-transitory machine-readable medium of claim 17 , wherein the determined emission directions include azimuth angles.
20 . The non-transitory machine-readable medium of claim 17 , wherein the determined emission directions include elevation angles.