| Summary: | 碩士 === 國立臺灣師範大學 === 資訊工程研究所 === 99 === We propose two transmission rate scheduling algorithms for sending time-sensitive multicast streaming data in a single hop wireless network. The algorithms are designed to allow the transmission of each sequence of encoded packets from a stream fragment to terminate in a shorter stopping time.
Assume that there exists a coding method for encoding forward-link packet streams such that it is sufficient for a user to decode a transmitted message by collecting k encoded packets. Also assume that the scheduler has perfect information of each channel state. Based on the assumptions, our approach is simply to choose for current time slot a peak feasible transmission rate, in terms of packets per slot. The transmission rate is chosen to achieve high multi-user diversity gain and high broadcast gain simultaneously, as developed in traditional approaches for large k. Besides, it is important for the chosen rate to limit the range of difference of packets received by each group user particularly when k is small. In this regards, we exploit the concept of grouping and arrange those users having received relatively less numbers of packets into a subgroup, called the least received-information subgroup whose size changes with time. We additionally consider the weight of received information lag among group users. We then develop two methods to choose the peak feasible rate from users in the subgroup. Specifically, one of them is to select the rate in the subgroup that maximizes the weighted volume of transmission to the least received-information subgroup; The other is to select the rate in the subgroup that maximizes the weighted volume of transmission to the whole group. The spirit under the two methods is that “simply raise the service priority of least received-information users because other users ahead in received information have more time to receive service opportunities.” Our results show that for small k and iid channel conditions, the proposed methods do achieve a shorter stopping time as compared to the best traditional approach designed for large k. This effect becomes more obvious for high channel SNR.
|
|---|
