查看文章

In packet-based wireless networks, media-based services often require a multicast-enabled transport that guarantees quasi error free transmission under strict delay constraints. Furthermore, both multicast and delay constraints deeply influence the architecture of erasure error recovery (EER). Therefore, we propose a general architecture of EER and study its optimization in this thesis. The architecture integrates overall existing important EER techniques: Automatic Repeat Request, Forward Error Correction and Hybrid ARQ techniques. Each of these EER techniques can be viewed as a special case of Hybrid Error Correction (HEC) schemes. Since the Gilbert-Elliott (GE) erasure error model has been proven to be valid for a wide range of packet based wireless networks, in this thesis, we present the general architecture and its optimization based on the GE channel model. The optimization target is to satisfy a certain target packet loss level under strict delay constraints efficiently. Through the optimization for a given real-time multicast scenario, the total needed redundancy information can be minimized by choosing the best HEC scheme automatically among the entire schemes included in the architecture. As a result, the performance of the optimum HEC scheme can approach the Shannon limit as closely as possible dynamically according to current channel state information.