User agents, systems and methods for machine-learning aided autonomous mobile network access
AssigneeCable Television Laboratories, Inc.
InventorsSayandev Mukherjee, Bernardo Huberman, Thomas Sandholm
Life of this patent
Estimating the term…
Abstract
Agents, operating on user equipment, and related systems and methods for machine-learning aided autonomous control of mobile network access are disclosed. The agents, systems and methods allow for mobile network resources, which are pre-purchased or purchased in real-time, to be optimized according to a reward function that maximizes quality of experience as a function of price. Mobile network resources purchased in real-time may include spectrum bandwidth sold or auctioned in increments and purchased by UEs as needed.
Claims11
An agent for autonomously controlling mobile network access, the agent comprising:
- an activity monitor, located on a user equipment (UE), that identifies an application accessed by a user;
- and
- a selector, located on the UE, that:
- (i) receives information from the activity monitor;
- (ii) determines resource requirements of the application;
- (iii) identifies a plurality of resource packages;
- (iv) predicts which one resource package of the plurality of resource packages meets the resource requirements and optimizes a reward function using a machine learning model trained on previously implemented resource packages and associated quality of experience (QoE) metrics;
- and
- (v) instructs a component of the UE to implement the one resource package.
Dependencies are inferred from the claim text.