On making channel output feedback practical for communication systems

Mayur Agrawal, Purdue University

Abstract

With the explosion of data, all communication systems need to provide increased data rates and higher reliability. This work investigates the advantages of using channel output feedback in communication systems to achieve these goals. We begin by looking at the role channel output feedback can play for additive Gaussian noise channels. In particular, we consider two scenarios, one with a point-point link and the other with a helper relay node. For a point-point link, we propose an optimal coding strategy that takes feedback noise into account. For a three terminal relay, we propose a new lower bound on the achievable rate in the case of ideal channel output feedback. We also propose a coding strategy for the noisy channel output feedback case. Subsequently, we explore the improvements in reliability for fading systems using channel output feedback. Most of the previous work for fading channels with feedback has been limited to the exploitation of channel state information (CSI) at the source. In this work, the design of a coding scheme for multiple-input multiple-output (MIMO) fading systems with channel output and channel state feedback is considered. We propose a simple linear coding strategy that achieves any rate up to the capacity of the channel. This linear processing scheme can provide a doubly exponential probability of error decay with the size of blocklength for any rate less than capacity. This scheme is extended to non-ideal feedback involving the following two scenarios: quantized CSI sent from the destination to the source and noisy channel output feedback from the destination to the source. Finally, we investigate practical performance benefits from our proposed scheme by incorporating it in the hybrid automatic repeat request (ARQ) protocol. Using simulations, we demonstrate that substantial improvements in the overall throughput of hybrid ARQ systems can be obtained by the adoption of our proposed scheme.

Degree

Ph.D.

Advisors

Balakrishnan, Purdue University.

Subject Area

Electrical engineering

COinS