Joint network coding and backpressure algorithm for cognitive radio networks

This paper presents the joint design of network coding and backpressure algorithm for cognitive radio networks and its implementation with software-defined radios (SDRs) in a high fidelity network emulation testbed. The backpressure algorithm is known to provide throughput optimal solutions to joint routing and scheduling for dynamic packet traffic. This solution applies to cognitive radio networks with spectrum dynamics changing over time and space, and supports joint routing and spectrum access without any need for end-to-end path maintenance. The backpressure algorithm is extended to multicast traffic with network coding deployed over virtual queues that represent different flows per session and destination. This extension is supported with different methods to decode packets at destinations. In the absence of a common control channel, distributed coordination with local information exchange is used to support neighborhood discovery, spectrum sensing and channel estimation that are integrated with joint routing, channel access and network coding. Cognitive network functionalities are implemented with GNU Radio and Python modules for different network layers, and used with USRP N210 radios. Practical radio implementation issues are addressed in a distributed wireless network setting, where USRP N210 radios communicate with each other through RFnest, a high fidelity wireless network emulation tool. RFnest provides realistic physical channel environment by digitally controlling path loss, fading, delay, topology and mobility effects. Extensive emulation test results are provided to assess throughput, backlog and energy consumption and verify the SDR implementation of joint network coding and backpressure algorithm under realistic channel and radio hardware effects.

I. Introduction

Spectrum is a scarce resource that is both time and location-dependent. Therefore, spectrum access should be jointly designed with routing decisions. The backpressure algorithm achieves the optimal throughput (within the stability region) for dynamic traffic by jointly optimizing routing and scheduling decisions [1]. There have been recent efforts to implement the backpressure algorithm on real radios such as in general multi-hop networks [2] and cognitive radio networks [3], [4]. The classical setting for the backpressure algorithm is unicast traffic (single source-destination pair per flow) whereas extension to multicast traffic (multiple destinations per flow) typically requires setting up and maintaining multi-hop tree structures [5]. Such an extension would be a challenging task in a distributed multi-hop wireless network, especially in a cognitive radio network with fast dynamics to be considered in routing and spectrum access decisions.