Machine learning assisted operations control
As an aspect, there is provided an apparatus, caused at least to: receive, by a user device from an access node, a measurement configuration for training of a machine learning operations control model, the measurement configuration comprising an indication of at least one parameter to be measured and at least one acceptance condition for measurement values; obtain at least one measurement value; compare the at least one measurement value to the at least one acceptance condition for defining usability of the at least one measurement value for the training, and when the at least one measurement value is usable for the training, carry out the training of the operations control model, and when the at least one measurement value is not usable for the training, transmit the at least one measurement value to the access node.
1. An apparatus comprising:
at least one processor and at least one memory including a computer program code, the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus at least to:
receive, by a user device from an access node, a measurement configuration for training of a machine learning operations control model, the measurement configuration comprising an indication of at least one parameter to be measured and at least one acceptance condition for measurement values of the at least one measured parameter, wherein the machine learning operations control model comprises a machine learning radio resource management model and machine learning internal operations model;
obtain at least one measurement value;
compare the at least one measurement value to the at least one acceptance condition for defining usability of the at least one measurement value for the training, and when the at least one measurement value is usable for the training,
carry out the training of the machine learning radio resource management model, and transmit the machine learning radio resource management model as trained to the access node for a radio resource management operation associated with the machine learning radio resource management model, and
carry out the training of the machine learning operations control model, wherein the machine learning radio resource management model as trained is used by the user device in the radio resource management operation associated with the machine learning radio resource management model and the machine learning internal operations model is used by the user device in user device operations associated with the internal operations model, and
when the at least one measurement value is not usable for the training,
transmit the at least one measurement value to the access node,
wherein the measurement configuration is area-specific, user device-specific, user device group-specific and service-specific.
2. The apparatus of claim 1 , wherein the obtaining, the comparing and the carrying out the training or the transmitting the at least one measurement value are repeated until the training of the operations control model reaches a target accuracy defined by at least one parameter associated to variation among the at least one measurement value.
3. The apparatus according to claim 1 , wherein the obtaining, the comparing and the carrying out the training or the transmitting the at least one measurement value are repeated until the training of the operations control model reaches a target accuracy defined by at least one parameter associated to variation among the at least one measurement value, further comprising causing the apparatus to indicate end of the repeating to the access node.
4. The apparatus according to claim 1 , wherein the obtaining, the comparing and the carrying out the training or the transmitting the at least one measurement value are triggered periodically and wherein the period is indicated in the measurement configuration.
5. The apparatus of claim 1 , wherein the measurement configuration comprises a radio resource control (RRC) state for which the configuration is to be valid for user device in connected, idle, or inactive state or for any combination of the states a user device may be in.
6. The apparatus of claim 1 , wherein when the user device is in connected state, the user device transmits the machine learning radio resource management model as trained to the access node when requested by the access node or by user device's own initiative.
7. The apparatus of claim 1 , wherein when the user device is in idle or inactive state, the user device transmits the machine learning radio resource management model as trained or the at least one measurement value to the access node after state change to connected state.
8. A method comprising:
receiving, by a user device from an access node, a measurement configuration, for training of a machine learning operations control model, the measurement configuration comprising an indication of at least one parameter to be measured and at least one acceptance condition for measurement values of the at least one measured parameter, wherein the machine learning operations control model comprises a machine learning radio resource management model and machine learning internal operations model;
obtaining at least one measurement value;
comparing the at least one measurement value to the at least one acceptance condition for defining usability of the at least one measurement value for the training, and when the at least one measurement value is usable for the training, carrying out the training of the machine learning radio resource management model, and transmitting the machine learning radio resource management model as trained to the access node for a radio resource management operation associated with the machine learning radio resource management model, and
carrying out the training of the machine learning operations control model, wherein the machine learning radio resource management model as trained is used by the user device in the radio resource management operation associated with the machine learning radio resource management model and the machine learning internal operations model is used by the user device in user device operations associated with the internal operations model, or
when the at least one measurement value is not usable for the training,
transmitting the at least one measurement value to the access node,
wherein the measurement configuration is area-specific, user device-specific, user device group-specific and service-specific.
9. The method of claim 8 , wherein the obtaining, the comparing and the carrying out the training or the transmitting the at least one measurement value are repeated until the training of the operations model reaches a target accuracy defined by at least one parameter associated to variation among the at least one measurement value.
10. The method of claim 8 , wherein the obtaining, the comparing and the carrying out the training or the transmitting the at least one measurement value are repeated until the training of the operations control model reaches a target accuracy defined by at least one parameter associated to variation among the at least one measurement value, further comprising indicating end of the repeating to the access node.
11. The method according to claim 8 , wherein the obtaining, the comparing and the carrying out the training or the transmitting the at least one measurement value are triggered periodically and wherein the period is indicated in the measurement configuration.
12. The method according to claim 8 , wherein there are a plurality of task-specific algorithms or models stored in the user device and the user device selects the one indicated in the measurement configuration, wherein the measurement configuration is adapted to a specific task, wherein the specific task is related to a radio resource management operation, wherein the radio resource management operation comprises at least one of handover, cell selection or cell reselection, adding a new component carrier, configuring or reconfiguring carrier aggregation, configuring or reconfiguring dual connectivity.
13. The method according to claim 8 , wherein there are a plurality of task-specific algorithms or models stored in the user device and the user device selects the one indicated in the measurement configuration, wherein the measurement configuration is adapted to a specific task, wherein the specific task is related to optimization of the user device's internal functions, wherein the internal functions comprise the utilization of its memory, battery resources, choice of antenna in beamforming operations, tuning and correction of the accuracy of positioning system and wherein when the user device is in connected state, the user device transmits the machine learning radio resource management model as trained to the access node when requested by the access node or by user device's own initiative.
14. The method according to claim 8 , wherein the measurement configuration comprises a timer for periodically triggering the training, wherein the timer monitors time from the end of a previous training period and wherein when the user device is in idle or inactive state, the user device transmits the machine learning radio resource management model as trained or the at least one measurement value to the access node after state change to connected state.
15. The method according to claim 8 , wherein the at least one measurement value is placed in a cluster in which the measurement value has a minimum distance from the mean value of all existing measurement values in the cluster, and when a measurement exceeds a threshold distance from the mean value of all the points in the cluster, it is considered to deviate from the machine learning model and is not used for the training, wherein the threshold distance is given in the measurement configuration.
16. The method according to claim 8 , wherein the at least one measurement value is placed in a cluster in which the measurement value has a minimum distance to a boundary point of the cluster, and when the measurement exceeds a threshold distance from the boundary point, it is considered to deviate from the machine learning model and is not used for the training, wherein the boundary point and the threshold distance are given in the measurement configuration.
17. The method according to claim 16 , wherein during the training, a cluster head is changed to a new one that minimizes the distance from all members of a new cluster created by adding one or more new measurement values to the model or algorithm until a target accuracy is reached, wherein the model is considered converged when the cluster head does not change when a new RSRP measurement is input to the model during a certain period or the new cluster head is near the previous one, wherein this period is indicated in the measurement configuration together with a threshold value for the acceptable distance to the previous cluster head.
18. The method according to claim 8 , wherein the measurement value is given a time stamp, the deviation from the machine learning model in terms of a mean square error, and location information.
19. The method according to claim 18 , wherein the obtaining, the comparing and the carrying out the training or the transmitting the at least one measurement value is triggered periodically, wherein the period is indicated in the measurement configuration, wherein the triggering is based on expiration of the time stamp is given in the measurement configuration.