Method and apparatus for transmitting data on basis of polar coding in wireless communication system
A method for transmitting data on the basis of polar coding in a wireless communication system, according to the present disclosure may comprise the steps of: transmitting data including a plurality of information blocks, each of which contains a corresponding cyclic redundancy check (CRC); receiving a hybrid automatic repeat request acknowledgement/negative acknowledgement (HARQ ACK/NACK) of the transmitted data; performing learning in order to retransmit the plurality of information blocks; and retransmitting the plurality of information blocks on the basis of the HARQ ACK/NACK, wherein the step of performing learning comprises the steps of: obtaining current state s n ; obtaining actions to be applied to current state s n ; and selecting, from the actions, an action maximizing expected reward value Q n+1 , wherein expected reward value Q n+1 is obtained on the basis of rewards R 1 , R 2 , . . . , R n corresponding to states s 1 , s 2 , . . . , s n , and the plurality of information blocks is retransmitted on the basis of the selected action.
1. A method of transmitting data in a wireless communication system, the method comprising:
dividing an information sequence into a plurality of information blocks, each having a size N b ;
adding a corresponding cyclic redundancy check (CRC) to each of the plurality of information blocks to produce a plurality of CRC added information blocks;
applying channel coding to each of the plurality of CRC added information blocks to produce a plurality of coded information blocks;
performing a first transmission including the plurality of coded information blocks;
receiving a first hybrid automatic repeat request acknowledgement/negative acknowledgement (HARQ ACK/NACK) for the first transmission;
selecting an action from among a set A of actions based on the first HARQ ACK/NACK and machine learning, wherein the actions in the set A are respectively related to transmission schemes; and
performing a second transmission based on the selected action,
wherein the machine learning comprises:
selecting an action A k for k-th transmission at a state S k ,
wherein the state S k =(k, U k−1 ), where k∈{1, 2, . . . , J}, where J is the maximum number of allowed retransmissions, and U k−1 is the sum of numbers of all coded bits transmitted up to (k−1)-th transmission;
obtaining a reward R k+1 and a state S k+1 =(k+1, U k ),
wherein R k+1 is determined based on N b *ACK k /U k , where ACK k is 1 for decoding success for the action A k , and ACK is 0 for decoding failure for the action A k ; and
updating an expected reward value Q based on i) an expected reward value Q for the state S k and the action A k , ii) the reward R k+1 and iii) α, where α is a learning rate and is a value in a range between 0 and 1.
2. The method of claim 1 , wherein the learning rate α is determined based on a range of channel variation.
3. The method of claim 1 , wherein the actions comprises:
a first action of transmitting all the plurality of information blocks which are not coded;
a second action of transmitting all the plurality of coded information blocks; and
a third action of transmitting coded information blocks for some information blocks among the plurality of information blocks and uncoded information blocks for remaining information blocks among the plurality of information blocks.
4. An apparatus for transmitting data based on in a wireless communication system, the apparatus comprising:
a transceiver;
at least one processor; and
at least one memory storing instructions that, when executed, cause the at least one processor to perform operations comprising:
dividing an information sequence into a plurality of information blocks, each having a size N b ;
adding a corresponding cyclic redundancy check (CRC) to each of the plurality of information blocks to produce a plurality of CRC added information blocks;
applying channel coding to each of the plurality of CRC added information blocks to produce a plurality of coded information blocks;
performing a first transmission including the plurality of coded information blocks;
receiving a first hybrid automatic repeat request acknowledgement/negative acknowledgement (HARQ ACK/NACK) for the first transmission;
selecting an action from among a set A of actions based on the first HARQ ACK/NACK and machine learning, wherein the actions in the set A are respectively related to transmission schemes; and
performing a second transmission based on the selected action,
wherein the machine learning comprises:
selecting an action A k for k-th transmission at a state S k ,
wherein the state S k =(k, U k−1 ), where k∈{1, 2, . . . , J}, where J is the maximum number of allowed retransmissions, and U k−1 is the sum of numbers of all coded bits transmitted up to (k−1)-th transmission;
obtaining a reward R k+1 and a state S k+1 =(k+1, U k ),
wherein R k+1 is determined based on N b *ACK/U k , where ACK k is 1 for decoding success for the action A k , and ACK k is 0 for decoding failure for the action A k ; and
updating an expected reward value Q based on i) an expected reward value Q for the state S k and the action A k , ii) the reward R k+1 and iii) α, where α is a learning rate and is a value in a range between 0 and 1.
5. The apparatus of claim 4 , wherein the learning rate α is determined based on a range of channel variation.
6. The apparatus of claim 4 , wherein the actions comprises:
a first action of transmitting all the plurality of information blocks which are not coded;
a second action of transmitting all the plurality of coded information blocks; and
a third action of transmitting coded information blocks for some information blocks among the plurality of information blocks and uncoded information blocks for remaining information blocks among the plurality of information blocks.
7. The method of claim 1 , wherein an action ‘a’ that maximizes an expected reward value Q for the state S k and the action ‘a’ is selected as A k with probability 1−ε, where a∈A, and A k is randomly selected from the set A with probability ε, where ε is related to a degree of exploration and exploitation and is a value in a range between 0 and 1.
8. The method of claim 7 , further comprising:
decreasing ε as the machine learning progresses, based on channel coherence time of a channel being stable.
9. The method of claim 8 , wherein decreasing ε as the machine learning progresses comprises:
decreasing ε at a first rate based on the channel having small channel coherence time; and
decreasing ε at a second rate lower than the first rate based on the channel having large channel coherence time.
10. The method of claim 8 , further comprising:
increasing ε based on the channel coherence time of the channel varying.
11. The method of claim 2 , wherein the learning rate α is proportional to 1/Var(|h|), where Var(|h|) is the range of channel variation.
12. The apparatus of claim 4 , wherein an action ‘a’ that maximizes an expected reward value Q for the state S k and the action ‘a’ is selected as A k with probability 1−ε, where a∈A, and A k is randomly selected from the set A with probability ε, where ε is related to a degree of exploration and exploitation and is a value in a range between 0 and 1.
13. The apparatus of claim 12 , wherein the operations further comprise:
decreasing ε as the machine learning progresses, based on channel coherence time of a channel being stable.
14. The apparatus of claim 13 , wherein decreasing ε as the machine learning progresses comprises:
decreasing ε at a first rate based on the channel having small channel coherence time; and
decreasing ε at a second rate lower than the first rate based on the channel having large channel coherence time.
15. The apparatus of claim 13 , wherein the operations further comprise:
increasing ε based on the channel coherence time of the channel varying.
16. The apparatus of claim 5 , wherein the learning rate α is proportional to 1/Var(|h|), where Var(|h|) is the range of channel variation.