Improved Spinal Codes: A Segmented CRC-Aided Scheme

As a new type of rateless codes, Spinal codes can be proved in theory that it can achieve capacity over both the additive white Gaussian noise (AWGN) channel and the binary symmetric channel (BSC) with short message length. Due to the good adaptability under different channel conditions, Spinal codes have broad prospects in ultra-reliable low-latency communication (URLLC) scenarios such as self-driving car and factory automation. However, Spinal codes transmitted by short codes need frequent times of feedback, while transmitted by long codes have a high decoding complexity, which limits the practical application of Spinal codes. In this work, a new type of encoding scheme named as segmented CRC-aided scheme is proposed. In this scheme, message is equally divided into λ segments, each of which is concatenated with a cyclic redundancy check (CRC) sequence. At the decoding end, all the segments are decoded in parallel, and the correspondingly CRC check results are collected and transmitted back to the encoder together. The encoder judges the current decoding state through the feedback and then constructs the next encoding pass accordingly. The segmented CRC-aided scheme combines the advantages of long codes transmission and short codes transmission of the Spinal codes, it uses fewer feedbacks, and it can reduce the transmission of redundance bits. Results demonstrate that the proposed scheme has significant performance improvement over the original Spinal encoding scheme by achieving higher code rate with lower encoding complexity.