Incremental redundancy hybrid automatic repeat request(IR HARQ) has been extensively studied for reliable data transmission over slow-fading or quasi-static channels.With the increase in movement speed of users and the use of long code words for data transmission,IR HARQ strategy in fast-fading channels is starting to attract attention in the academia.This paper studies the performance of the IR HARQ strategy based on Kite codes(a class of rateless codes) in the finite regime over fast-fading channels where a number of channel realizations are experienced in each retransmission round.We propose an algorithm that exploits current decoding reliability to determine the size of subsequent retransmissions.Longterm throughput and delay constraint throughput are analyzed and compared.Furthermore,in HARQ systems available,most of the computation power is consumed on failed decoding if a code word is retransmitted many times,which is not energy-efficient.Therefore,to improve theenergy efficiency,we propose two efficient algorithms(early stopping algorithm and freezing node algorithm) for incremental decoding,which reduce the computational complexity of the most time-consuming steps in decoding procedure.Simulation results show that the substantial complexity reduction is achieved in terms of the total required number of decoding iterations and the required node operation complexity compared to conventional incremental decoding scheme.
