LEARNING-BASED SEMI-PERSISTENT SCHEDULING IN WIRELESS COMMUNICATIONS

摘要

Systems and methods are provided for a learning-based determination of se mi-persistent scheduling of data-packet flow wireless communication. A packe tized data flow served to a wireless terminal is fully scheduled for an init ial period of time in order to collect statistics associated with scheduled packet sizes (Ss) and inter-packet times (Ts). Analysis of a cumulative dist ribution of {S, T} pairs indicate whether a characteristic packet size (S0) and size dispersion (D0) are associated with the cumulative distribution. In ter-time intervals associated with the characteristic size and dispersion co mplete a transport format. Semi-persistent scheduling is utilized for a pack etized flow when a characteristic transport format can be extracted, or lear ned, from the accumulated statistics. Extracted transport formats can be emp loyed to optimize scheduling efficiency upon handover.