Rank metric convolutional codes for Random Linear Network Coding

Random Linear Network Coding (RLNC) currently attracts a lot of attention as a technique to disseminate information in a network. In this contribution, non-coherent multi-shot RLNC is considered, that means, the unknown and time variant network is used several times. In order to create dependencies between the different shots, convolutional network codes are used, in particular Partial Unit Memory (PUM) codes. Such PUM codes based on rank metric block codes are constructed and it is shown how they can efficiently be decoded when errors, erasures and deviations occur. The decoding complexity of this algorithm is cubic with the length. Further, it is described how lifting of these codes can be applied for error correction in RLNC.