System and method for providing multiple input multiple output (MIMO) detector selection with reinforced learning neural network
View Patent ↗A method and system for selecting a symbol detector are herein provided. A method includes extracting a first set of features for a k-th resource element (RE), where k is an integer greater than one, extracting a second set of features from a first RE to a (k−1)th RE, and selecting a symbol detector for the k-th RE using a reinforcement learning (RL) neural network based on the extracted first set of features and the extracted second set of features.
1. A method for selecting a symbol detector, comprising:
extracting a first set of features for a k-th resource element (RE), where k is an integer greater than one;
extracting a second set of features from a first RE to a (k−1)th RE; and
selecting a symbol detector for the k-th RE using a reinforcement learning (RL) neural network based on the extracted first set of features and the extracted second set of features.
2. The method of claim 1 , wherein the first set of features are based on a channel matrix for the k-th RE.
3. The method of claim 1 , wherein the second set of features are based on accumulated log likelihood ratios (LLR).
4. The method of claim 1 , wherein the second set of features include a normalized location, absolute values of log likelihood ratio (LLR) distributions or soft symbol distribution.
5. The method of claim 1 , wherein the RL neural network includes a multi-layer perceptron (MLP).
6. The method of claim 1 , wherein the RL neural network generates a complexity score corresponding to a complexity of the symbol detector.
7. The method of claim 6 , wherein the symbol detector is selected based on the generated complexity score.
8. The method of claim 1 , wherein the RL neural network generates a decoding penalty indicating whether decoding will be successful.
9. The method of claim 8 , wherein the decoding penalty is based on a Lagrange multiplier penalty parameter.
10. The method of claim 1 , wherein the RL neural network is trained using a deep Q-network (DQN).
11. A system for selecting a symbol detector, comprising:
a memory; and
a processor configured to:
extract a first set of features for a k-th resource element (RE), where k is an integer greater than one;
extract a second set of features from a first RE to a (k−1)th RE; and
select a symbol detector for the k-th RE using a reinforcement learning (RL) neural network based on the extracted first set of features and the extracted second set of features.
12. The system of claim 11 , wherein the first set of features are based on a channel matrix for the k-th RE.
13. The system of claim 11 , wherein the second set of features are based on accumulated log likelihood rations (LLR).
14. The system of claim 11 , wherein the second set of features include a normalized location, absolute values of log likelihood ratio (LLR) distributions or soft symbol distribution.
15. The system of claim 11 , wherein the RL neural network includes a multi-layer perceptron (MLP).
16. The system of claim 11 , wherein the RL neural network generates a complexity score corresponding to a complexity of the symbol detector.
17. The system of claim 16 , wherein the symbol detector is selected based on the generated complexity score.
18. The system of claim 11 , wherein the RL neural network generates a decoding penalty indicating whether decoding will be successful.
19. The system of claim 18 , wherein the decoding penalty is based on a Lagrange multiplier penalty parameter.
20. The system of claim 11 , wherein the RL neural network is trained using a deep Q-network (DQN).