Feedback in wireless networks cross-layer design, secrecy and reliability /

by 1981- Gopala, Praveen Kumar

The central theme of this dissertation is the impact of feedback on the performance of wireless networks. Wireless channels offer a multitude of new challenges and opportunities that are uncharacteristic of wireline systems. We reveal the crucial role of feedback in exploiting the opportunities and in overcoming the challenges posed by the wireless medium. In particular, we consider three distinct scenarios and demonstrate the different ways in which feedback helps improve performance. We first consider cellular multicast channels and show that the availability of feedback allows for the cross-layer design of efficient multicast schedulers. Here we focus on two types of feedback scenarios: perfect channel state information (CSI) feedback and automatic repeat request (ARQ) feedback. We propose low-complexity multicast schedulers that achieve near-optimal asymptotic throughput-delay tradeoffs for both feedback scenarios. We further propose a cooperative multicast scheduler, requiring perfect CSI feedback, that achieves the optimal asymptotic scaling of both throughput and delay with the number of users. For the multiple transmit antenna scenario, we show that the wireless multicast gain dominates the throughput performance of multicast schedulers and demonstrate the near-optimality of the proposed worst user scheduler with a large number of transmit antennas. Next, we consider fading eavesdropper channels and reveal the importance of feedback in establishing secure communications. We characterize the secrecy capacity ii of such channels under the assumptions of full CSI and main (legitimate) channel CSI knowledge at the transmitter, and propose optimal rate and power allocation strategies. Interestingly, we show that the availability of CSI feedback enables one to exploit the time-varying nature of the wireless medium and achieve a perfectly secure non-zero rate even when the eavesdropper channel is more capable than the legitimate receiver channel on the average. We further establish the critical role of rate adaptation, based on the main channel CSI, in facilitating secure communications over slow fading channels. We also propose a low-complexity on/off power allocation strategy and establish its asymptotic optimality. We then consider a minimal ARQ feedback scenario and propose transmission schemes that leverage the ARQ feedback to achieve non-zero perfect secrecy rates even when the eavesdropper has a superior channel on the average. Thereby, we establish the positive impact of feedback on the secrecy capacity of fading channels. Finally, we consider ARQ channels with strict delay deadline constraints and demonstrate the impact of ARQ feedback on reliability. We show that ARQ feedback can be exploited to significantly improve the achievable error exponents, and propose an Incremental Redundancy ARQ (IR-ARQ) scheme that significantly outperforms the schemes based on memoryless decoding. iii To Amma, Achan, and Chinnu, for their love and support. iv