Resource-efficient Mechanisms for Wireless Coexistence
The convergence of networked embedded devices, wearables, and sensing technologies has expedited the emergence of an array of innovative services and applications that are radically changing the way we perceive and interact with the physical world. Wireless communication is the natural substrate connectivity means for a wide variety of these applications. For these applications to perform correctly, they require the underlying wireless communication to be reliable and energy-efficient. Meeting these requirements is, however, challenging. Particularly, as we witness an unprecedented demand for wireless access, more wireless technologies and devices need to share the scarcely available radio spectrum. This is especially a growing problem for devices operating in the unlicensed spectrum, where the density and heterogeneity of radios operating in this spectrum are surging. Consequently, interference between the heterogeneous radio systems is growing in unpredictable ways. The emerging spectrum crunch necessitates the design and development of innovative wireless systems that enhance spectrum utilization and are apprehensive of the uncoordinated wireless coexistence problem.

In this dissertation, we take an alternative approach to deal with Cross-Technology Interference (CTI). Instead of avoiding interference, we adopt an interdisciplinary approach combining a cross-layer design and machine learning techniques to build cognitive low-power wireless systems that can cope with Cross-Technology Interference. We begin this dissertation by acquiring a good understanding of how various interfering wireless signals interact, and we harness this understanding in our designs. We then introduce a family of algorithms and system architectures that improve the robustness of low-power wireless networks operating in interference-rich environments. The introduced systems embody a cross-layer design and a cognitive engine that radios can exploit to intelligently share the spectrum and implement CTI-aware mitigation schemes. In particular, we present three novel systems contributing to low-power wireless systems coexistence:

i) Technology-IndependentInterferenceMitigation(TIIM): Interfering radio technologies di↵er widely in the way they affect wireless links. Cross-Technology Interference has a complex impact on wireless links, which needs to be taken into account when treating interference.



Resource-efficient Mechanisms for Wireless Coexistence
To address this challenge, we present TIIM, a system that identifies, quantifies, and reacts to CTI in real-time. In the design of TIIM, we follow an unconventional approach, where we employ lightweight machine learning techniques to assist wireless nodes in recovering from interference. Within TIIM, we develop a lightweight classifier which is trained to select a coexistence solution that works most e↵ectively for the current channel fingerprint.

ii) CrossZig: Current wireless designs still largely impose layer isolation. Thereby, conventional approaches to tackle wireless performance have focused on separately optimizing different layers of the networking stack. This rigid design fails to harness the rich ambient information embedded in the physical signals. Hence, reliability solutions targeting layers in isolation are typically suboptimal. In recent years, cross-layer optimizations were profoundly advocated in the wireless community. In this work, we pursue this research direction. We show how physical layer information and primitives can be coupled with the link layer to enhance low-power wireless systems coexistence and performance under interference. Notably, we present CrossZig, a cross-layer wireless design, that enables low-power wireless networks to exploit fine-grained physical layer information to make informed decisions that can help them recover from varying sources of interference. CrossZig utilizes physical layer information to detect the presence of external interference in corrupted packets and to apply an adaptive packet recovery which incorporates a novel cross-layer based packet merging scheme and an adaptive channel coding.

iii) Controlled Interference Generator (CIG): Wireless research testbed infrastructures often lack proper tools for enabling repeatable replay of realistic radio interference commonly found in real-world deployments. Hence, benchmarking wireless coexistence solutions is often cumbersome, time-consuming, and even infeasible in remote testbeds. To facilitate Cross-Technology Interference and wireless coexistence experimentations, we develop CIG, a framework that extends wireless testbed infrastructures with the capability of reproducing heterogeneous external interference. In the design of CIG, we consider a unified approach that incorporates a careful selection of interferer technologies (implemented in software), to expose networks to realistic interference patterns.

The systems presented in this dissertation demonstrate that incorporating cognitive and cross-layer wireless designs is adequate to mitigate the problem of uncoordinated wireless coexistence.