Research On The Fair Resource Scheduling In Wireless Networks

This thesis studies the fair scheduling problem in wireless networks. For threetypes of typical networks, we study the problem that in wireless fading environments,how to improve the resource usage efficiency to take advantage of the limited systemresources while guaranteeing the system fairness. For different designing objective, thisthesis proposes different types of fair scheduling strategies accordingly. First, we studythe queue-stability oriented fair scheduling schemes in an opportunistic relay networks.Then for a generalized multi-base-station multi-user network, we propose two types ofscheduling schemes that both achieve the proportional fairness in a network-wide sense.Finally for the cognitive networks, we consider the problem of joint channel probingand user scheduling. The major contributions are given in the following.Firstly, we consider an opportunistic relay netwok that allows user cooperation.The objective is to maintain all users’ queue stability. Based on the instantaneouschannel state information and queue buffer information, we propose two algorithms thatselect both the desitination user and the relaying user. The obtained algorithms expandthe system’s queue-stable throughput region, reduce the average queue delay andimprove the fairness indexes of both the user throughput and queuing delay.Secondly, we study the network-wide proportional fair scheduling problem in amulit-base-station multi-user network under the deterministic channel rate assumption.An online pricing algorithm is proposed, which achieve network-wide proportionalfairness with the interaction between base staions and users. The distributedimplementation of the algorithm is given. And the convergence and optimality of thealgorithm is proved mathematically. Simulations show that the proposed algorithmouterperform the traditional local proportional fair scheduling algorithm.Thirdly, we study the global proportional fair scheduling problem in amulit-base-station multi-user network under the stochastic channel rate assumption.Using the stochastic approximation method, we propose a scheduling scheme withbase-station cooperation, which achieves the global proportional fairness. Theasymptotic optimality of this scheme is proved. Furthermore, we use a group of ordinary differentional equations (ODE) to characterize all links’ dynamic throughputtrajectories. We propose two types of implementations of the scheme, one for networksthat allows inter-base-station communication and one for network whereinter-base-station communication is unfeasible. For both the implementations,complexity analysis is given. We show that compared with loca proportional fairscheduling scheme, our scheme improves both the throughput and fairness.Fourthly, we study the impact of the user channel probing on the fair scheduling.The problem is modeled as an optimal stopping problem under the assumption that thescheduler only knows user-channels’ statistics, and we design a joint channel probingand user schemdling algorithm. We prove the algorithm’s convergence, optimality andproportional fairness. Also a static-threshold stopping criteria is proposed that reducesthe algorithm complexity substantially. And we derive the explicit formula for thealgorhtm’s multi-user diversity gain. Furthermore, the algorithm is extended to tha casewhere the scheduler does not know user-channels’ statictics at all. We propose a jointonline learning, channel probing and user scheduling algorithm and show itsconvergance to the standard algorithm.