Resource allocation based on hierarchical and nonlinear control in wireless networks

With the ever increasing popularity of wireless data services and applications, allocation of radio resources such as bandwidth and power is extremely important in wireless networks which carry multiple traffic types with different quality-of-service (QoS) requirements. In this thesis, we consider resource allocation for direct-sequence code division multiple access (DS-CDMA) systems that carry circuit-switched as well as packet-switched traffic. We apply classical techniques from nonlinear control and hierarchical control theory to design and analyze a class of dynamic power and rate control algorithms that can optimize the overall network performance under constraints on frame error rate, delay and transmit power.; We first apply nonlinear control techniques to analyze a quantized transmit power control algorithm that is commonly used for both voice and data services. Modeling a CDMA system with power control as a nonlinear feedback control system, we apply classical control techniques such as phase-plane and describing functions to study the stability and convergence of power control. We then apply a statistical linearization method to analyze power control performance in a fading environment. The resulting approach provides a framework for analyzing the effect of design parameters such as power control step size, mobile speed and fading channel characteristics on power control performance.; Next, we use hierarchical control techniques to study combined power and rate control schemes on the downlink for data users. We consider a class of data retransmission protocols such as the Radio Link Protocol (RLP) where an errored frame is retransmitted for a limited number of times. We first consider a circuit-switched data system where data arrives at a constant rate which can be controlled by the network. The QoS for data users is specified by a family of utility functions that trades off throughput and fairness among the data users. The network optimizes the sum of utility functions of all users while maintaining certain constraints on transmit power, delay and the resulting frame error after the RLP. We view this as a static large scale system and derive an optimal control algorithm with an hierarchical structure that can be implemented in a distributed manner. Specifically, the data rates for the users are adjusted (jointly) by the base station while the powers are controlled (individually) by the mobiles.; Finally, we consider the downlink of a packet-switched system where data arrives into a data queuing buffer according to a certain arrival process but the average arrival rate can be controlled by the network. We consider optimization of the sum of utility functions of all data users under constraints on frame error rate, average and peak transmit power as well as delay. We view this as a dynamic large-scale system and derive the optimal control law for rate allocation as well as dynamic buffer management (framing). We show that the overall control algorithm has an hierarchical structure where the individual controllers adjust the rate and power for each user dynamically and a centralized controller guarantees that the power and delay constraints are met for all users. Further, we show that by adjusting certain design parameters in the utility functions, a flexible tradeoff between the overall optimality and fairness among the users can be achieved.