Toward Robust and Context-aware Wireless In-band Full-duplex Networks

A few decades ago, the concept of smart cities and smart factories could only be pictured in science fiction movies. However, the recent massive advances in computer, information and communication technologies have accelerated progress toward realizing this concept. We literally cannot imagine such a landscape without a robust wireless network over which one can communicate reliably. This has urged the researchers to push the technology boundaries aiming for Ultra-Reliable and Low-Latency Communication (URLLC) where services have crucial obligations for availability and latency. Many applications quickly come to mind, namely, Vehicular Ad-hoc Network (VANET), Vehicular Edge Computing, Industry 4.0, and Industry 5.0 which can already be seen on the horizon. This PhD work is an attempt to improve the reliability of wireless networks and provide them with environmental sensing capability. As the key enabler, we make use of In-band Full-duplex (IBFD) technology that allows for simultaneous transmission and reception over the same frequency band, and enables constant monitoring of the wireless spectrum. IBFD is recognized for the potential to double the throughput of a communication link, but in fact, it is not its only benefit. On the one hand, such a technique allows a communication unit to recapture the environmental reflections of what it transmits, which opportunistically can be used by an integrated radar to render environment probing. On the other hand, it enables the device to listen to the channel while it transmits data, so that it can detect a likely signal interference and take an appropriate reaction. This kind of channel context sensing improves network reliability, specifically in the contention-based channel sharing schemes. Using IBFD as an opportunity, this dissertation investigates four techniques to facilitate context-awareness and signal collision detection; two approaches for each item are explored. First, I introduce a communication transceiver design in which a mono-static Doppler-radar is also integrated. The design benefits from an analog module to provide sufficient transmitter-receiver isolation, so that the radar can reuse the communication signal to sense the environmental Dopplers. I develop an analytical system model to study different influential parameters that impact the Doppler sensing performance. This context-sensing approach is also prototyped over a real-time platform. The analytical model is then validated by numerical simulation as well as measurements through the prototype. In the second step, a joint IBFD communication and radar processing scheme is introduced, which also achieves optimal hardware resource utilization. Based on a mathematical system model and simulation, I implement a prototype and analyze the communication-radar performance in a real-world bi-directional communication link. Finally, I study two distinct approaches to implement instantaneous signal collision detection to improve network reliability. To this end, a convolutional neural network is employed to compete with a model that is a design based on noise and distortion estimation. Then, a measurement-based simulation evaluates each approach performance, taking into account the detection probability and the false alarm rate as two critical metrics. This PhD thesis concludes that the IBFD technology not only can improve the network throughput but also offers the opportunity to achieve environmental and spectral context-awareness. The experimental results show that the opportunistic radar solution can detect concurrent targets at 10 and 20m by reusing the self-transmit communication signal that carries data to a second party node. The developed technique is cost-efficient as it can be implemented by a small set of additional logic components. Furthermore, the proposed collision detection strategies in this thesis outperform the current state-of-the-art technique by a 12-fold faster detection, while they still have low-complexity in terms of real-time realization. More specifically, the results show that collisions up to 30dB below the reference signal (the self-interference) can be detected precisely within 0.02ms at the cost of negligible probability of false alarm.

Publication year:2021

Accessibility:Closed