Research And Modification Of Wireless Packet Scheduling Algorithm Based On Proportion Fairness

Future mobile communication system needs to support higher transmission rate, accommodate more users, and protect the quality of service (QoS) for a variety of businesses. In a multi-user shared wireless network, to design effective wireless packet scheduling algorithm is helpful to achieving this goal. In wireless networks, the effectiveness and fairness of scheduling technology are usually two conflicting performances. Scheduling algorithms must obtain the best compromise between these two indicators based on QoS requirements of different users. Therefore, to design a packet scheduling algorithm which not only can ensure user's QoS, but also improve the system capacity is currently a hot research topic in wireless communication field.By analyzing the characteristics of wireless networks, the situation of wireless packet scheduling technology is introduced, the classification, design objectives and performance evaluation criteria are given, and thesis points out some of the challenges facing now. Firstly, the packet scheduling mechanism in time-division multiplexing wireless network with time-varying environment is studied. Taking into account the diversity of business needs of future users, discussion takes data business and voice business as example, respectively. Because different types of business have their own characteristics and different QoS requirements, the considerations are different during scheduling algorithm design. For example, data business often has a large amount of data need to transport, which require higher data rate from scheduling algorithms to ensure the system throughput. Voice business is very sensitive to delay, requiring algorithms to minimize delay and ensure short-term fairness.The thesis first describes the packet scheduling model in TDMA system, introduces RR, MAX C/I and PF three classic algorithms applied to data business. Taking into account the factor of the user's access delay, in order to improve the user's short-term fairness, modified proportion fairness (M-PF) algorithm is proposed based on PF algorithm. This algorithm maintains an access delay timer for each user, so that the user whose access delay is close to the maximum delay can get prior scheduling opportunity. Simulation results show that, M-PF algorithm is strict fair on long-term, as well as greatly improves the short-term fairness of the system, only in terms of system throughput has very little loss than PF algorithm. For voice business studies, the two classic scheduling algorithms M-LWDF and EXP are analyzed. Taking into account specific requirements (packet timeout rate, timeout limit, etc.) for packet delay of voice business, still using the idea of PF, weighted exponent proportion fairness (WEPF) algorithm is proposed. The algorithm adds exponentiated packet waiting time into discriminative rule, sets a prioritizing parameter, and time delay compensation mechanism is introduced by set a pre-timeout threshold, so priority of nearly timeout packet can be raised, the possibility of being scheduled will be increased. Under the assumption of same and different user's QoS the simulation results show that the WEPF algorithm greatly reduces user's packet delay and jitter, improves the system's overall performance and better ensures the quality of service.As a core technology of next generation mobile communication, OFDM technology can increase spectrum efficiency and improve system's ability of resistance to multipath fading. Therefore, packet scheduling algorithm in OFDM system has become nowadays a hot topic. Since resources in OFDM systems are two-dimensional time-frequency, so the design of scheduling and resource allocation scheme for this system is different from systems based on time-division or code-division multiplexing.Finally, wireless packet scheduling algorithm in the OFDM system is studied. First briefly introduces the OFDM technology, describes the resource allocation scheme in OFDM system, an OFDM system scheduling model is given. Then a mathematical model of the system and three scheduling algorithms (K&H, McPF and K&H with sub-carrier fine-tuning) for the model are given. Performances of different algorithms are compared. Simulation results show that in the cost of small system throughput loss, McPF algorithm achieves good compromise in terms of fairness and throughput. At the same time, adding sub-carriers fine tuning can better ensure the performance of short-term fairness.