| Summary: | 博士 === 國立成功大學 === 電腦與通信工程研究所 === 100 === This dissertation examines two issues of multiple-group communications in spatial time division multiple access (STDMA) networks. Firstly, STDMA networks should provide an effective solution for enabling wireless devices to access network resources with fairness and efficiency. When multiple-group communications are implemented in such networks, the scheduling algorithm should generate appropriate schedule assignments for all the transmissions where the objective aims to reduce the schedule length. In this dissertation, the problem of producing an efficient schedule sequence for multiple-group communications over a STDMA scheduling network is referred to as an integrated multiple-group communication and traffic-oriented scheduling (IMCTS) problem. It is shown that the IMCTS problem can be formulated as an integer linear programming (ILP) problem. Two polynomial-time link- and node-based scheduling algorithms, designated as source-parallel-aware link assignment (SLA) and source-parallel-aware node assignment (SNA), respectively, are proposed for determining the schedule sequence subject to transmission constraints. In addition, this dissertation shows that a effect of bottleneck path and a inappropriate interference model impact on the transmission performance of traffic-oriented scheduling networks. To enhance the spatial utilization efficiency within each time slot, an advanced version of SNA, designated as collision-allowed source-parallel-aware node assignment (CSNA), is proposed based on a modified graph-based interference model. An advanced version of SLA, designated as joint source-parallel-aware link assignment with network coding (JSLANC), is proposed to minimize the effects on bottleneck paths on the schedule frame length by flexibly applying conventional or opportunistic network coding approaches. It is shown that compared to existing TDMA- and STDMA-based algorithms, the proposed algorithms provide an effective reduction in the schedule frame length and a significant increase in the spatial utilization within each time slot.
Secondly, STDMA networks provide a local loss recovery for ensuring reliable group communications. In local loss recovery schemes, a small number of recovery nodes distributed along the transmission paths save incoming packets temporarily in accordance with a specified cache policy and retransmit these packets if they subsequently receive a request message from a downstream receiver. To reduce the recovery latency, the cache policy should ensure that the recovery nodes are always able to satisfy the retransmission requests of the downstream receivers. However, due to the limited cache size of the recovery nodes and the behavior of the cache policy, this cannot always be achieved, and thus some of the packets must be retransmitted by the sender. Accordingly, this dissertation develops a new network-coding-based cache policy, designated as network-coding-based FIFO (NCFIFO), which extends the caching time of the packets at the recovery nodes without dropping any of the incoming packets. As a result, the lost packets can be always recovered from the nearest recovery nodes and the recovery latency is significantly reduced. The loss recovery performance of the NCFIFO cache policy is compared with that of existing cache policies by performing a series of simulation experiments using both a uniform error model and a burst error model. The simulation results show that the NCFIFO cache policy not only achieves a better recovery performance than existing cache policies, but also provides a more effective solution for managing a small amount of cache size in environments characterized by a high packet arrival rate.
|
|---|
