摘要： In this paper we develop Prioritized Automatic Code Embedding (PACE) link-layer protocol to achieve preferred data recovery order across connections, while maintaining stable and reliable data flows over wireless networks. We classify link-layer traffic arrivals into different priorities based on the packet delay constraint and the distortion associated with the loss of that packet. The traffic arriving in each priority class is modeled as a poisson process. Consequently, we formulate the link-layer buffer as a multiclass M/G/1 priority queuing system where the decoding process (service process) of the PACE buffer is captured by a nonhomogeneous geometric distribution. This formulation enables the determination of an optimal dynamic decoding schedule for heterogeneous link-layer traffic arrivals. PACE employs a novel rate-adaptive Low Density Parity Check (LDPC) channel code for error recovery. We model the underlying LDPC decoding process and use it to determine an optimal code selection policy for maximal bandwidth utilization. We demonstrate experimentally that PACE reduces the throughput-delay cost by 20%-70% in comparison with the IEEE802.11 ARQ and Hybrid ARQ (HARQ) protocols. Further, the PACE protocol achieves 20%-60% improvement in channel bandwidth utilization and 2-10dB PSNR gain in realtime video quality.
