| Summary: | 碩士 === 國立臺灣大學 === 資訊管理學研究所 === 92 === With the rise of the Internet, now many data transfer applications are essential to people’s daily life. In addition, wireless networks can support people’s mobility to access information regardless of where they are. Hence mobile communication has become a popular topic in today’s technology world.
Due to the rapid growth of the wireless LAN market, the need of transmitting real-time and multimedia traffic, such as voice, images, video and …, etc, over wireless LAN will gradually increase. Therefore, the relevant Quality of Service (QoS) problem has also become a critical issue. However, as there are some inherent differences between wireless networks and traditional wired networks. So with this pre-determined condition, such as using CSMA/CA in its MAC protocol under IEEE 802.11, we will need more and more effective mechanism to provide QoS assurance.
In this thesis, we bring up the concept of slotting the contention-free period in IEEE 802.11 WLAN. In addition, to be compatible with IEEE 802.11e, we consider four data frame types with different priorities. As a result, in the limited wireless spectrum resource, different slot allocation policy will generate results varying in revenue, throughput and QoS of the system. However, there is always a tradeoff between the users’ QoS requirement and the system revenue. Therefore, we hope to provide differentiated service while maximizing the long-term system revenue.
We propose two mathematical models to solve the slot allocation problem in this thesis. The goal of our model is to find the best slot allocation policy to maximize the long-term system revenue under the capacity constraint. The main difference is the time type. The first model is continuous-time, while the second one is discrete-time. We apply Markovian Decision Process to deal with our problem due to the problem size and the structure of our model.
According to the good computational results, we can successfully find the best slot allocation policy that maximizes the long-term system profit by Markovian Decision Process. Compared with the heuristics that venders often use, the policy we find has great improvement in the system revenue. Therefore, our model can indeed provide good decision for system venders and network planners.
|
|---|
