Optimization of hybrid ARQ error control schemes

Abstract

If an error occurs due to insufficient information in a received signal, a receiver can request a retransmission by feeding back control signals. This is called the Automatic Repeat and reQuest (ARQ) error control scheme. When error correction control is also combined together, we call it Hybrid-ARQ (HARQ). If the receiver combines the previously received signal which failed to be decoded with a newly received signal, it can utilize more information than the ARQ schemes. This is the main concept of Type-II Hybrid ARQ. There are two types of HARQ schemes: Chase Combining (CC) type HARQ which transmits the same packet for each retransmission, and Incremental Redundancy (IR) type HARQ which transmits new information for each retransmission. The CC-type HARQ schemes increase reliability and throughput by obtaining temporal diversity and power gains. The IR-type HARQ schemes improve the performance more by additionally obtaining coding gain through the new information than the CC-type HARQ. However, it is much harder to implement the IR-type HARQ than the CC-type HARQ scheme due to encoding and decoding complexity.

HARQ schemes are effective particulary when a transmitter cannot accurately expect the successful decoding. For example, when the channel gain fluctuates rapidly but the transmitter cannot obtain the exact channel gain feedback, the error probability severely increases, compared with invariant channel. Even though the channel gain does not fluctuate in AWGN channels, if the length of packet is finite, the error probability is not arbitrarily small even for a coding rate less than the channel capacity. The motivation of this dissertation is to maximize the long-term average transmission rate (LATR) of HARQ schemes under an uncertain channel condition with the limitation on the maximum number of (re)transmissions and an error probability constraint. One important control parameter is the coding rate which represents the link adaptation.

In the first p...