System and method for joint power and resource allocation using reinforcement learning
View Patent ↗Systems and methods for joint power and resource allocation on a shared 5G channel. The method selects one of a group of grouped actions and implements this selected group of actions. The effects of the actions on the environment and/or the users are then assessed. Based on the result, a reward is allocated for the system. Multiple iterations are then executed with a view to maximizing the reward. Each of the grouped actions comprises joint power and resource allocation actions.
1. A method for managing resources for a shared channel being shared between users that are classified under different service categories, the method comprising:
(a) providing at least one software agent in a gNodeB;
(b) causing said at least one software agent to select one action group to be executed for at least one user from a service category;
(c) executing said action group for said at least one user from said service category;
(d) assessing at least one effect of said action group on a measurable metric affecting said at least one user;
(e) determining a reward result based on said at least one effect of said action group;
(f) repeating steps (b)-(e) to thereby maximize said reward result over multiple iterations of steps (b)-(e);
wherein said one action group is selected from a predetermined group of action groups, each of said action groups including actions that relate to at least one of: power allocations and resource block allocations;
said at least one software agent comprises a plurality of software agents; there is an absence of cooperation between said plurality of software agents; and said absence of cooperation between said plurality of software agents is such that action group selections are not communicated between each of said plurality of software agents.
2. The method according to claim 1 , wherein said shared channel is a 5G channel.
3. The method according to claim 2 , wherein said service category is Ultra-Reliable Low-Latency (URLLC).
4. The method according to claim 2 , wherein said service category is enhanced Mobile Broadband (eMBB).
5. The method according to claim 1 , wherein each of said software agents observes a state of said shared channel to execute step d).
6. The method according to claim 5 , wherein said state is based on a signal to interference plus noise ratio value for said at least one user.
7. The method according to claim 5 , wherein said state is based on an average estimate of a signal to interference plus noise ratio value users classified under a specific category.
8. The method according to claim 1 , wherein execution of each action group results in joint power and resource block allocations for a given scheduled interval.
9. The method according to claim 1 , wherein execution of each action group results in power allocations for a given scheduled interval.
10. The method according to claim 1 , wherein execution of each action group results in resource block allocations for a given scheduled interval.
11. The method according to claim 1 , wherein said measurable metric is at least one of: user latency, reliability, transmission delay, queueing delay, and scheduling delay.
12. The method according to claim 1 , wherein said reward result is based on a packet queueing delay for said at least one user.
13. The method according to claim 1 , wherein said reward result is based on an average throughput for users classified under a specific category.
14. The method according to claim 1 , wherein said reward result is based on a packet queueing delay for said at least one user classified under a first specific category and on an average throughput for users classified under a second specific category.
15. A system for managing resources for a shared channel, said shared channel being shared between users that are classified under different service categories, the system comprising a processor configured to execute:
a link adaptation module for allocating resources to modules processing packets for multiple users, each of said multiple users being classified under different categories and for determining at least one measurable metric for said shared channel;
a scheduler module receiving said at least one measurable metric from said link adaptation module; wherein said scheduler module executes a method comprising:
causing at least one software agent to select one action group to be executed for at least one user from one of said different service categories;
executing said action group for said at least one user from said service category to thereby allocate resources to said modules by way of said link adaptation module;
assessing at least one effect of said action group on said at least one measurable metric affecting said at least one user;
determining a reward result based on said at least one effect of said action group;
repeating said method to thereby maximize said reward result over multiple iterations of said method;
wherein said one action group is selected from a predetermined group of action groups, each of said action groups including actions that relate to at least one of: power allocations and resource block allocations;
said at least one software agent comprises a plurality of software agents; there is an absence of cooperation between said plurality of software agents; and said absence of cooperation between said plurality of software agents is such that action group selections are not communicated between each of said plurality of software agents.