加权比例公平群智能跨层资源分配算法

Weighted swarm intelligence cross-layer resource allocation algorithm with proportional fairness

针对多用户OFDM系统,提出两种适用于混合业务的加权比例公平跨层资源分配方案。该方案假设系统用户拥有多个队列,每个队列分别承载不同类型的业务。在MAC层,所提的两种方案都实施加权比例公平调度。该调度先为用户队列中不同分组授予不同的权重,再通过该权重值计算用户权重,并对每个用户的分组进行排序,最后根据系统中各用户待传数据量之比设置用户间速率成比例约束条件。在物理层,这两种方案不仅都将用户间速率成比例约束条件下系统权重容量和的最大化作为优化目标,而且都在该目标下将群智能算法引入其资源分配。但有所不同的是,方案1将人工鱼群算法引入其子载波分配,用新推导的功率分配方式进行功率分配;方案2将云自适应粒子群算法引入其子载波分配,用人口迁移算法进行功率分配。在此基础上,两种方案都依据由加权比例公平调度提供的各用户分组排序结果传送分组。数值仿真与性能分析显示,这两种方案能在满足用户业务流时延需求和保证用户公平性的基础上,有效提高系统总速率。

This paper proposed two weighted cross-layer resource allocation schemes with proportional fairness for multi-user OFDM system on the basis of researching the swarm ifitelllgence algorithm, which were suitable for heterogeneous traffic. It not only assumed that every user in the system had multiple queues, that every queue carried respectively traffic of different types. In the MAC layer, the proposed two schemes both carried out the weighted proportional fairness scheduling. The scheduling provided different packets in users＇ queues different weight, sorted every user＇ s packets calculating users＇ weight by means of the weight, and set up the proportional constraints among users＇ rates according to the ratio of users＇ data to be passed. In the PHY layer, the two schemes not only both defined maximizing the system weighted capacity sum as optimization target under the proportional constraints among users＇ rates, but also both led＇ swarm intelligence algorithm into their resource allocation under the optimization target. However, scenario 1 led artificial fish swarm algorithm into its sub-carrier allocation and used a new power allocation pattern to carry out power allocation, scenario 2 led the cloud adaptive particle swarm optimization algo- rithm into its sub-carrier allocation and used population migration algorithm to carry out power allocation. At last, the two schemes transmitted packets according to packet sequencing results that provided by the weighted proportional fairness scheduling. Data simulation and performance analysis demonstrate that the two schemes can increase the total system rates effectively on the basis of meeting the delay of users＇ traffic and guaranteeing users＇ fairness.