| Efficient resource management is one of the key components in the design of wireless communication systems. There is plenty of literature on cross-layer resource control to improve physical layer performance, such as throughput, signal to interference and noise ratio (SINR) and bit error rate (BER), etc. The derived resource control polices are adaptive to the channel state information (CSI) only. A typical assumption in these papers is that the transmitters have infinite backlogs and the information flows are delay-insensitive. In practice, a lot of applications in wireless communications systems are delay-sensitive. Therefore, it is also very important to consider random bursty arrivals and delay performance in resource control. This thesis takes a cross-layer approach to address delay-optimal resource control problems to minimize average delay for delay-sensitive applications in wireless communication systems. The derived resource control polices are adaptive to both the CSI and the queue state information (QSI).;We consider the delay-optimal resource control design and delay performance analysis for tandem networks under different link connectivity conditions. We borrow the calculus approach and use queueing theory as well as dynamic programming in continuous-time optimal control to obtain the delay-optimal and asymptotical optimal resource control designs. Using renewal theory, theory of random walks and sample path analysis, we analyze the delay performance of the delay-optimal design and the delay penalty of the throughput-optimal dynamic backpressure (DBP) policy compared with the delay-optimal policy.;We consider the delay-optimal resource control designs in single-cell orthogonal frequency-division multiple access (OFDMA) systems and cooperative and coordinated multiple-input and multiple-output (MIMO) systems in cellular networks. We formulate the corresponding delay-optimal resource control problems as infinite horizon average cost per stage Markov decision processes (MDPs). To obtain low complexity and distributed delay-aware resource control designs, we carefully approximate the potential functions and Q-factors using approximate MDP and estimate them using distributed stochastic learning. We show the proposed low complexity distributed algorithms converge almost surely. In addition, we also analyze the asymptotic delay performance of the proposed solution or compare the delay performance of the proposed solution with the corresponding base policies. |
