Wireless communication systems and methods using federated learning and supervised preferential optimization
Aspects of the subject disclosure may include, for example, systems and methods for improving wireless communication by selecting best configurations of intelligent reflecting surface (IRS) devices present in different wireless communication environments with supervised preferential optimization (SPO) technique and sharing parameters of the best configurations of IRS devices with a global shared model under federated learning setting. Other embodiments are disclosed.
1 . A system, comprising:
a plurality of wireless communication network components implementing different wireless communication environments, wherein, in each of the plurality of wireless communication network components, one or more base stations and one or more user equipment (UE) communicate via an intelligent reflecting services (IRS) device present therein and wherein the IRS device tracks a different wireless communication environment, resulting in a different configuration of the IRS device;
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:
communicating with each of the plurality of wireless communication network components in a federated learning (FL) setting;
maintaining a global shared model having gold standard parameters of a configuration of the IRS device;
receiving, from each of the plurality of wireless communication network components, a parameter set, wherein the parameter set represents a best configuration of the IRS device based on the different wireless communication environment thereof, wherein the best configuration of the IRS device is learned and selected using a supervised preference optimization (SPO) machine learning technique; and
updating the gold standard parameters of the global shared model based on the received parameter set.
2 . The system of claim 1 , wherein the operations further comprise sending updated parameters of the global shared model to the plurality of wireless communication network components via a feedback loop.
3 . The system of claim 1 , wherein the operations further comprise aggregating the parameter set from each of the plurality of wireless communication network components to form unified parameters.
4 . The system of claim 1 , wherein the best configuration of the IRS device is selected based on expert preference data and artificial intelligence/machine learning (AI/ML) decision-making policies.
5 . The system of claim 4 , wherein each of the plurality of wireless communication network components comprises an AI/ML model configured to capture the different wireless communication environment thereof and select the best configuration of the IRS device based on the captured different wireless communication environment and the expert preference data customized to the captured different wireless communication environment.
6 . The system of claim 5 , wherein the AI/ML decision-making policies comprise a preferred configuration of the IRS device and a not preferred configuration of the IRS device based on the expert preference data.
7 . The system of claim 6 , wherein the expert preference data comprise high strength signal and minimum interference during a predetermined time frame,
the preferred configuration of the IRS device comprises a configuration of the IRS device that reflects signals toward high-density areas in the captured different wireless communication environment, and
the not preferred configuration of the IRS device comprises a configuration of the IRS device that reflects signals toward low-density areas in the captured different wireless communication environment.
8 . The system of claim 6 , wherein the expert preference data comprise seamlessly maintaining reliable connectivity for one or more selected areas,
the preferred configuration of the IRS device comprises a configuration of the IRS device that constantly directs signals toward the one or more selected areas in the captured different wireless communication environment, and
the not preferred configuration of the IRS device comprises a configuration of the IRS device that directs signals toward other areas than the one or more selected areas in the captured different wireless communication environment.
9 . 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:
maintaining a shared intelligent reflecting surface (IRS) configuration model;
communicating with a plurality of wireless communication network components configured to represent different wireless communication environments, wherein each wireless communication network component comprises an action selection module running a supervised preference optimization (SPO) model and wherein the communicating comprises communicating with the action selection module in each of the plurality of wireless communication network components in a federated learning (FL) setting;
receiving, from the action selection module, a parameter set, wherein the parameter set include information indicating a selected configuration of an IRS device operating in and tracking the different wireless communication environment, wherein the selected configuration of the IRS represents a best configuration that has been learned and determined using the SPO model;
aggregating the parameter set from each action selection module of each wireless communication network component in the FL setting; and
updating the shared IRS configuration model based on aggregated parameter sets.
10 . The non-transitory machine-readable medium of claim 9 , wherein the operations further comprise sending a feedback based on the updated shared IRS configuration model to the action selection module of each wireless communication network component.
11 . The non-transitory machine-readable medium of claim 9 , wherein the parameter set does not include operation data and performance data of each wireless communication network component, wherein the operation data and the performance data are locally maintained in each wireless communication network component.
12 . The non-transitory machine-readable medium of claim 9 , wherein the best configuration of the IRS device is selected based on expert preference data and artificial intelligence/machine learning (AI/ML) decision-making policies, and
wherein the action selection module is configured to capture a different wireless communication environment thereof and select the best configuration of the IRS device based on the captured different wireless communication environment and the expert preference data customized to the captured different wireless communication environment.
13 . The non-transitory machine-readable medium of claim 12 , wherein the AI/ML decision-making policies comprise a preferred configuration of the IRS device and a not preferred configuration of the IRS device based on the expert preference data.
14 . The non-transitory machine-readable medium of claim 13 , wherein the expert preference data comprise high strength signal and minimum interference during a predetermined time frame,
the preferred configuration of the IRS device comprises a configuration of the IRS device that reflects signals toward high-density areas in the captured different wireless communication environment, and
the not preferred configuration of the IRS device comprises a configuration of the IRS device that reflects signals toward low-density areas in the captured different wireless communication environment,
wherein the best configuration of the IRS device is selected to adopt the preferred configuration of the IRS device and avoid the not preferred configuration of the IRS device.
15 . The non-transitory machine-readable medium of claim 13 , wherein the expert preference data comprise seamlessly maintaining reliable connectivity for one or more selected areas,
the preferred configuration of the IRS device comprises a configuration of the IRS device that constantly directs signals toward the one or more selected areas in the captured different wireless communication environment, and
the not preferred configuration of the IRS device comprises a configuration of the IRS device that directs signals toward other areas than the one or more selected areas in the captured different wireless communication environment,
wherein the best configuration of the IRS device is selected to adopt the preferred configuration of the IRS device and avoid the not preferred configuration of the IRS device.
16 . A method, comprising:
maintaining, by a processing system including a processor, a shared intelligent reflecting surface (IRS) configuration model;
communicating, by the processing system, with a plurality of wireless communication network components configured to implement different wireless communication environments, wherein each wireless communication network component comprises an IRS device including an action selection module, and wherein the action selection module is configured to adopt a supervised preference optimization (SPO) model and wherein the communicating comprises communicating with the action selection module in each of the plurality of wireless communication network components in a federated learning (FL) setting;
receiving, by the processing system, from the action selection module, a parameter set, wherein the parameter set include information indicating selected IRS configuration operating in and tracking the different wireless communication environment, wherein the selected IRS configuration represents a best configuration that has been learned and determined using the SPO model;
aggregating, by the processing system, the parameter set from each action selection module of each wireless communication network component in the FL setting; and
updating, by the processing system, the shared IRS configuration model based on aggregated parameter sets.
17 . The method of claim 16 , comprising:
sending, by the processing system, via a feedback loop, updated parameters of the shared IRS configuration model to the action selection module of each wireless communication network component.
18 . The method of claim 16 , wherein the best configuration of the IRS device is selected based on expert preference data and artificial intelligence/machine learning (AI/ML) decision-making policies, and
wherein the action selection module is configured to capture a different wireless communication environment thereof and select the best configuration of the IRS device based on the captured different wireless communication environment and the expert preference data customized to the captured different wireless communication environment.
19 . The method of claim 18 , wherein the AI/ML decision-making policies comprise a preferred configuration of the IRS device and a not preferred configuration of the IRS device based on the expert preference data based on the SPO model.
20 . The method of claim 19 , wherein the best configuration of the IRS device is selected to adopt the preferred configuration of the IRS device and avoid the not preferred configuration of the IRS device.