Balanced resource allocator for heterogeneous multi-objective systems
A computer-implemented method, a system, and a computer program product include program code that obtains, from a controller, temporal objectives and configuration specifications. The program code adjusts weighting for multiple concurrent modes. The program code optimizes each mode of the multiple concurrent modes. The program code discovers conflicts exist between the optimized multiple concurrent modes and resolves the conflicts. The program code allocates resources to the optimized multiple concurrent modes.
1. A computer-implemented method comprising:
obtaining, by one or more processors, from a controller, temporal objectives and configuration specifications;
adjusting, by the one or more processors, based on the objectives and configurations, weighting for multiple concurrent modes;
optimizing, by the one or more processors, each mode of the multiple concurrent modes;
discovering, by the one or more processors, conflicts exist between the optimized multiple concurrent modes and resolving the conflicts; and
allocating, by the one or more processors, resources to the optimized multiple concurrent modes.
2. The computer-implemented method of claim 1 , wherein the multiple concurrent modes comprise multiple concurrent radio frequency (RF) modes.
3. The computer-implemented method of claim 2 , wherein optimizing each mode of the multiple concurrent modes comprises:
computing, by the one or more processors, optimal resource allocation across multiple concurrent RF modes by utilizing a graph-based constrained shortest path approach.
4. The computer-implemented method of claim 3 , wherein utilizing the graph-based constrained shortest path approach algorithm comprises implementing a graph-based algorithm via dynamic programming.
5. The computer-implemented method of claim 1 , wherein optimizing each mode of the multiple concurrent modes comprises:
obtaining a performance prediction for each mode of the multiple concurrent modes by a learning engine communicatively coupled to the one or more processors.
6. The computer-implemented method of claim 5 , wherein obtaining the performance prediction comprises:
detecting, by the learning engine, a context of the temporal environment;
based on the context, determining, by the learning engine, whether to utilize a long time horizon or a short time horizon to utilize in the performance prediction for each for each of the multiple concurrent modes; and
formulating, by the one or more processors, the performance prediction based on the selection.
7. The computer-implemented method of claim 6 , wherein detecting the context comprises determining if the temporal environment is stationary or dynamic.
8. The computer-implemented method of claim 6 , wherein detecting the context comprises determining if the temporal mode has a priori model.
9. The computer-implemented method of claim 6 , wherein the determining whether to utilize the long time horizon or the short time horizon to utilize in the performance prediction comprises selecting a machine learning policy from the group consisting of: model-free reinforcement learning, a Markov decision process, a rule-based policy, and a single-step Markov decision process.
10. The computer-implemented method of claim 9 , wherein the model-free reinforcement learning machine learning policy is selected based on detecting the temporal environment is a stationary environment.
11. The computer-implemented method of claim 9 , wherein the model-free reinforcement learning machine learning policy is selected based on detecting the temporal environment comprises no a priori model.
12. The computer-implemented method of claim 9 , wherein the Markov decision process machine learning policy is selected based on detecting the temporal environment comprises a completed a priori model.
13. The computer-implemented method of claim 9 , wherein the rule-based policy machine learning policy is selected based on detecting the temporal environment is a dynamic environment.
14. The computer-implemented method of claim 9 , wherein the single-step Markov decision process machine learning policy is selected based on detecting the temporal environment is a dynamic environment.
15. The computer-implemented method of claim 9 , wherein the single-step Markov decision process machine learning policy is selected based on detecting the temporal environment comprises a completed a priori model.
16. The computer-implemented method of claim 1 , further comprising:
obtaining, by the one or more processors, feedback, based on the allocation; and
re-allocating, by the one or more processors, a portion of the resources, based on the feedback.
17. A system comprising:
a memory;
one or more processors in communication with the memory;
program instructions executable by the one or more processors via the memory to perform a method, the method comprising:
obtaining, by the one or more processors, from a controller, temporal environment objectives and configuration specifications;
adjusting, by the one or more processors, based on the objectives and configurations, weighting for multiple concurrent modes;
optimizing, by the one or more processors, each mode of the multiple concurrent modes;
discovering, by the one or more processors, conflicts exist between the optimized multiple concurrent modes and resolving the conflicts; and
allocating, by the one or more processors, resources to the optimized multiple concurrent modes, wherein the allocating enables the optimized multiple concurrent modes to change the occupied resources.
18. The system of claim 17 , wherein the optimizing is based on obtaining a performance prediction for each of the multiple concurrent modes by a learning engine communicatively coupled to the one or more processors.
19. The system of claim 18 , wherein obtaining the performance prediction comprises:
detecting, by the learning engine, a context of the temporal environment;
based on the context, determining, by the learning engine, whether to utilize a long time horizon or a short time horizon to utilize in the performance prediction for each for each of the multiple concurrent modes; and
formulating, by the one or more processors, the performance prediction based on the selection.
20. A computer program product comprising:
a non-transitory computer readable storage medium readable by one or more processors and storing instructions for execution by the one or more processors for performing a method comprising:
obtaining, by the one or more processors, from a controller, temporal environment objectives and configuration specifications;
adjusting, by the one or more processors, based on the objectives and configurations, weighting for multiple concurrent modes;
optimizing, by the one or more processors, each mode of the multiple concurrent modes;
discovering, by the one or more processors, conflicts exist between the optimized multiple concurrent modes and resolving the conflicts; and
allocating, by the one or more processors, resources to the optimized multiple concurrent modes, wherein the allocation enables the optimized multiple concurrent modes to change the occupied resources.