Low-Density Parity-Check (LDPC) codes are usually decoded by running aniterative belief-propagation, or message-passing, algorithm over the factorgraph of the code. The traditional message-passing schedule consists ofupdating all the variable nodes in the graph, using the same pre-updateinformation, followed by updating all the check nodes of the graph, again,using the same pre-update information. Recently several studies show thatsequential scheduling, in which messages are generated using the latestavailable information, significantly improves the convergence speed in terms ofnumber of iterations. Sequential scheduling raises the problem of finding thebest sequence of message updates. This paper presents practical schedulingstrategies that use the value of the messages in the graph to find the nextmessage to be updated. Simulation results show that these informed updatesequences require significantly fewer iterations than standard sequentialschedules. Furthermore, the paper shows that informed scheduling solves somestandard trapping set errors. Therefore, it also outperforms traditionalscheduling for a large numbers of iterations. Complexity and implementabilityissues are also addressed.
展开▼
