Reinforcement learning machine learning-assisted beam pair selection for handover in radio access networks
A system and method carried out over a mobile wireless network are described for performing beam pair (BP) and end-to-end (E2E) network slice selection for supporting an invoked service on a mobile equipment (ME). The method includes establishing an initial BP with a radio access network (RAN) node, using an available link policy, enabling communicating a request to the RAN node including an indication of a desired service level for a service invoked on the ME. The method further includes updating, in accordance with the indication of a desired service level, a link policy and an E2E slice policy by performing a reinforcement learning, wherein the link policy is used to select a BP for the ME for a given ME mobility pattern, and wherein the E2E network slice policy is used to select an E2E network slice for the desired service level for the service invoked on the ME.
1. A method carried out over a mobile wireless network for performing beam pair (BP) and end-to-end (E2E) network slice selection for supporting an invoked service on a mobile equipment (ME), the method comprising:
establishing an initial BP with a radio access network (RAN) node, using an available link policy, enabling communicating a request to the RAN node including an indication of a desired service level for a service invoked on the ME;
updating, in accordance with the indication of a desired service level, a link policy and an E2E network slice policy by performing a reinforcement learning, wherein the link policy is used to select a BP for the ME for a given ME mobility pattern, and wherein the E2E network slice policy is used to select an E2E network slice for the desired service level for the service invoked on the ME; and
selecting an E2E network slice including a target BP selected according to the link policy, to support the service invoked by the ME.
2. The method of claim 1 , wherein the indication of a desired service level comprises a key performance indicator (KPI) profile.
3. The method of claim 1 , wherein during the reinforcement learning, an action is specified to an environment by an agent, wherein the action proposes a changes to at least one of the group consisting of:
a previously proposed link policy; and
a previously proposed E2E network slice policy.
4. The method of claim 1 , further comprising providing the link policy to the RAN node.
5. The method of claim 1 , wherein during the updating, an exit criterion is specified for terminating performing the reinforcement learning.
6. The method of claim 5 , wherein the exit criterion comprises a quantity corresponding to a maximum number of episodes of an action/next state and reward specification cycle of the reinforcement learning.
7. The method of claim 1 , wherein the link policy is specified in a link policy table including entries identifying distinct mobility patterns of the ME and corresponding link performance indicator reward values.
8. The method of claim 7 wherein each distinct mobility pattern is specified by a geospatial location and velocity combination.
9. The method of claim 1 , wherein the E2E network slice policy is specified in a link policy table including entries identifying distinct E2E network slice patterns and corresponding service performance indicator reward values.
10. The method of claim 9 , wherein each distinct E2E network slice pattern is specified by a slice resource.
11. A networked system comprising:
a processor; and
a non-transitory computer-readable medium including computer-executable instructions that, when executed by the processor, facilitate carrying out a method carried out over a mobile wireless network for performing beam pair (BP) and end-to-end (E2E) network slice selection for supporting an invoked service on a mobile equipment (ME), the method comprising:
establishing an initial BP with a radio access network (RAN) node, using an available link policy, enabling communicating a request to the RAN node including an indication of a desired service level for a service invoked on the ME;
updating, in accordance with the indication of a desired service level, a link policy and an E2E network slice policy by performing a reinforcement learning, wherein the link policy is used to select a BP for the ME for a given ME mobility pattern, and wherein the E2E network slice policy is used to select an E2E network slice for the desired service level for the service invoked on the ME; and
selecting an E2E network slice including a target BP selected according to the link policy, to support the service invoked by the ME.
12. The system of claim 11 , wherein the indication of a desired service level comprises a key performance indicator (KPI) profile.
13. The system of claim 11 , wherein during the reinforcement learning, an action is specified to an environment by an agent, wherein the action proposes a changes to at least one of the group consisting of:
a previously proposed link policy; and
a previously proposed E2E network slice policy.
14. The system of claim 11 , further comprising providing the link policy to the RAN node.
15. The system of claim 11 , wherein during the updating, an exit criterion is specified for terminating performing the reinforcement learning.
16. The system of claim 15 , wherein the exit criterion comprises a quantity corresponding to a maximum number of episodes of an action/next state and reward specification cycle of the reinforcement learning.
17. The system of claim 11 , wherein the link policy is specified in a link policy table including entries identifying distinct mobility patterns of the ME and corresponding link performance indicator reward values.
18. The system of claim 17 wherein each distinct mobility pattern is specified by a geospatial location and velocity combination.
19. The system of claim 11 , wherein the E2E network slice policy is specified in a link policy table including entries identifying distinct E2E network slice patterns and corresponding service performance indicator reward values.
20. The system of claim 19 , wherein each distinct E2E network slice pattern is specified by a slice resource.