Abstract:

Four types of variable-rate convolutional network codes are investigated over a single-source finite cyclic network. It is found that variable-rate generic, dispersion and broadcast can be implemented on the same network without changing the local encoding kernels of the non-source nodes. The efficient implementation has the advantage that each non-source node only needs to store one copy of the local encoding kernel within a session. However, it is also shown by an example that variable-rate multicast may not always be implemented under the above condition.

Key words:

variable-rate, convolutional generic, convolutional dispersion, convolutional broadcast, convolutional multicast