Machine learning model selection for beam prediction for wireless networks
A method includes beam prediction inference using a first machine learning (ML) model based algorithm that uses a first quantity of reference signal (RS) measurements; transmitting an indication that the first user device is using the first ML model based algorithm that uses the first quantity of RS measurements; receiving, based on beam change dynamics event information received by the network node from one or more other user devices that have one or more corresponding conditions within a threshold to the first user device, a request for the first user device to either: change to a non-ML model based algorithm to perform beam selection or a second ML model based algorithm that uses a second quantity of reference signal measurements to perform beam prediction inference, wherein the second quantity is different than the first quantity, or concurrently perform beam prediction inferences using the first and second ML model based algorithms.
1 . An apparatus comprising: at least one processor and at least one memory including 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:
perform, by a first user device, beam prediction inference using a first machine learning model based algorithm that uses a first quantity of reference signal measurements;
control transmitting, by the first user device to a network node, an indication that the first user device is using the first machine learning model based algorithm that uses the first quantity of reference signal measurements;
control receiving, by the first user device from the network node based on beam change dynamics event information received by the network node from one or more other user devices that have one or more corresponding conditions within a threshold to the first user device, a request for the first user device to change to at least one of a non-machine learning model based algorithm to perform beam selection or a second machine learning model based algorithm that uses a second quantity of reference signal measurements to perform beam prediction inference, wherein the second quantity is different than the first quantity,
wherein the beam change dynamics event information comprises:
information indicating that a rate of change or frequency of change of a best beam for the first user device is either increasing or becoming more dynamic, or is decreasing or becoming less dynamic,
information indicating that a beam pattern for the first user device has changed,
information indicating one or more propagation environment changes associated with a change in beam selection for the first user device,
information indicating a request transmitted by the first user device to the network node for the first user device to change from the first machine learning model based algorithm that uses the first quantity of reference signal measurements to perform beam prediction inference to either the non-machine learning model based algorithm to perform beam selection or to a different machine learning model based algorithm to perform beam prediction that uses a quantity of reference signal measurements that is different than the first quantity of reference signal measurements;
information indicating that a performance of beam prediction inference of the first user device based on the first machine learning model based algorithm that uses the first quantity of reference signal measurements is less than a first threshold level of performance; and
information indicating that the performance of the first machine learning model based algorithm to perform beam prediction inference by the first user device is within a second threshold level of performance to a performance of either the non-machine learning model based algorithm or the different machine learning model based algorithm to perform beam prediction inference,
wherein the one or more corresponding conditions within the threshold for the first user device and one or more other user devices comprise:
the first user device is of a same device model or is manufactured by a same device vendor as the one or more other user devices;
the first user device has a location that is within a threshold distance to a location of the one or more other user devices;
there is a spatial correlation of a channel, within a first threshold, for the first user device and the one or more other user devices;
there is a correlation, within a second threshold, of reference signal measurements of the first user device and reference signal measurements of the one or more other user devices; and
the first user device and the one or more other user devices exhibit a same channel propagation characteristics, within a third threshold, including received power and phase shift, for received reference signals;
detect, by the first user device, beam change dynamics event information associated with a change in a beam selection for the first user device;
control transmitting, by the first user device to the network node, the beam change dynamics event information associated with a change in a beam selection for the first user device; and
control transmitting, by the first user device to the network node, information indicating a mapping between values of at least one field of a message and a plurality of algorithms that may be used by the first user device to perform beam prediction inference or beam selection.