Methods, apparatus and computer-readable mediums relating to detection of sleeping cells in a cellular network
The disclosure provides methods, apparatus and machine-readable mediums for the detection of sleeping cells in a cellular network. A method of detecting a sleeping cell in a cellular communication network comprises: monitoring power consumption of a radio transmission point of the cellular communication network; providing the power consumption of the radio transmission point as an input to a classification model, developed using a machine-learning algorithm; and obtaining an output from the classification model, the output classifying the radio transmission point as serving one of a sleeping cell and a non-sleeping cell.
1. A method of detecting a sleeping cell in a cellular communication network, the method comprising:
monitoring, by a radio transmission point, power consumption of the radio transmission point of the cellular communication network;
providing, by the radio transmission point, the power consumption of the radio transmission point as an input to a classification model located at the radio transmission point, developed using a machine-learning algorithm, wherein the classification model is a final classification model that is generated by amalgamating a plurality of locally trained classification models obtained from a plurality of radio transmission points; and
obtaining, by the radio transmission point, an output from the classification model, the output classifying the radio transmission point as serving one of a sleeping cell and a non-sleeping cell, wherein a power reset of the radio transmission point or a power amplifier of the radio transmission point is initiated in response to the radio transmission point being classified as a sleeping cell.
2. The method of claim 1 , wherein the classification model comprises a first part, developed using the machine-learning algorithm, for predicting a future power consumption pattern of the radio transmission point based on the monitored power consumption, and a second part for comparing the future power consumption to a measured power consumption and classifying the radio transmission point as serving one of a sleeping cell and a non-sleeping cell based on the comparison.
3. The method of claim 2 , wherein the second part of the classification model is also developed using a machine-learning algorithm.
4. The method of claim 1 , wherein the output of the classification model relates to a prediction that the radio transmission point will serve one of a sleeping cell and a non-sleeping cell.
5. The method of claim 1 , wherein the power consumption of the radio transmission point is monitored at a power amplifier of the radio transmission point.
6. The method of claim 5 , wherein the power consumption of the radio transmission point is monitored via one or more of a supply rail of the power amplifier and a biasing voltage rail of the power amplifier.
7. The method of claim 1 , wherein the power consumption of the radio transmission point is the only input to the classification model.
8. The method of claim 1 , wherein the power consumption is monitored as time-series data in a time window, and wherein the time-series data is provided as an input to the classification model.
9. The method of claim 8 , wherein the output of the classification model classifies the power consumption in the time window as relating to one of a sleeping cell and a non-sleeping cell.
10. The method of claim 1 , further comprising obtaining an output from the classification model periodically.
11. The method of claim 1 , wherein the machine-learning algorithm comprises a recurrent neural network.
12. The method of claim 1 , further comprising, responsive to a determination that the radio transmission point is serving a sleeping cell, initiating an action to remediate the sleeping cell.
13. The method of claim 12 , wherein the action comprises one or more of:
transmitting an indication of the sleeping cell to a scheduler of the cellular communication network;
resetting the radio transmission point; and
resetting a base station to which the radio transmission point belongs.
14. The method of claim 1 , further comprising updating the classification model based on the input and the output.
15. The method of claim 14 , further comprising transmitting the updated classification model to a remote server.
16. The method of claim 1 , further comprising receiving the classification model from a remote server.
17. A network node for detecting a sleeping cell in a cellular communication network, the network node comprising processing circuitry and a non-transitory computer-readable medium storing instructions which, when executed by the processing circuitry, cause the network node to:
monitor power consumption of a radio transmission point of the cellular communication network, wherein the network node includes the radio transmission point;
provide the power consumption of the radio transmission point as an input to a classification model, developed using a machine-learning algorithm, wherein the classification model is a final classification model that is generated by amalgamating a plurality of locally trained classification models obtained from a plurality of radio transmission points; and
obtain an output from the classification model, the output classifying the radio transmission point as serving one of a sleeping cell and a non-sleeping cell, wherein a power reset of the radio transmission point or a power amplifier of the radio transmission point is initiated in response to the radio transmission point being classified as a sleeping cell.
18. The network node of claim 17 , wherein the classification model comprises a first part, developed using the machine-learning algorithm, for predicting a future power consumption pattern of the radio transmission point based on the monitored power consumption, and a second part for comparing the future power consumption to a measured power consumption and classifying the radio transmission point as serving one of a sleeping cell and a non-sleeping cell based on the comparison.
19. The network node of claim 17 , wherein the network node is further caused to, responsive to a determination that the radio transmission point is serving a sleeping cell, initiate an action to remediate the sleeping cell.
20. A non-transitory computer-readable medium storing instructions which, when executed by processing circuitry of a network node of a cellular communication network, cause the network node to:
monitor power consumption of a radio transmission point of the cellular communication network, wherein the network node includes the radio transmission point;
provide the power consumption of the radio transmission point as an input to a classification model, developed using a machine-learning algorithm, wherein the classification model is a final classification model that is generated by amalgamating a plurality of locally trained classification models obtained from a plurality of radio transmission points; and
obtain an output from the classification model, the output classifying the radio transmission point as serving one of a sleeping cell and a non-sleeping cell, wherein a power reset of the radio transmission point or a power amplifier of the radio transmission point is initiated in response to the radio transmission point being classified as a sleeping cell.