Machine-learning assisted environment detection framework for self-adapting inter-RAT steering strategy

An apparatus implementable in a user equipment (UE), comprising:

• one or more sensors;
• a radio frequency (RF) circuit configured to communicate wirelessly;
• and
• a processor coupled to the one or more sensors and the RF circuit, the processor configured to perform operations comprising:
• extracting one or more feature metrics regarding a wireless network environment based at least partially on sensor information received from the one or more sensors and RF signal information from the RF circuit;
• identifying a scenario with respect to a current status of the wireless network environment according to the one or more feature metrics, wherein the scenario comprises a long-period out-of-service (OOS) or a short-period OOS;
• and
• performing, via the RF circuit, a radio access technology (RAT)-related operation responsive to the identifying of the scenario;
• wherein the RAT-related operation comprises holding on a call session in an event that the scenario is the short-period OOS,
• in extracting the one or more feature metrics, the processor is configured to perform operations comprising:
• training a machine-learning model with previous sensor information and previous RF signal information;
• and
• extracting the one or more feature metrics by utilizing the machine-learning model based on the received sensor information and RF signal information;
• in identifying the scenario, the processor is configured to perform operations comprising:
• performing environment classification on the extracted one or more feature metrics;
• determining that the scenario is a recurrent scenario according to a result of the environment classification;
• and
• updating the machine learning model with a result of the environment classification, as current status, via an on-device learning mechanism which involves:
• comparing feature similarity metrics to map the scenario to one or more recurrent scenarios;
• and
• updating a database with a result of the comparing,
• wherein the database is utilized by the machine-learning model in extracting the one or more feature metrics;
• in performing the RAT-related operation, the processor is configured to perform inter-RAT steering based on one or more RAT selection strategies according to the database.