Academic Commons

Theses Doctoral

Algorithms and Experimentation for Future Wireless Networks: From Internet-of-Things to Full-Duplex

Chen, Tingjun

Future and next-generation wireless networks are driven by the rapidly growing wireless traffic stemming from diverse services and applications, such as the Internet-of-Things (IoT), virtual reality, autonomous vehicles, and smart intersections. Many of these applications require massive connectivity between IoT devices as well as wireless access links with ultra-high bandwidth (Gbps or above) and ultra-low latency (10ms or less). Therefore, realizing the vision of future wireless networks requires significant research efforts across all layers of the network stack. In this thesis, we use a cross-layer approach and focus on several critical components of future wireless networks including IoT systems and full-duplex (FD) wireless, and on experimentation with advanced wireless technologies in the NSF PAWR COSMOS testbed.

First, we study tracking and monitoring applications in the IoT and focus on ultra-low-power energy harvesting networks. Based on realistic hardware characteristics, we design and optimize Panda, a centralized probabilistic protocol for maximizing the neighbor discovery rate between energy harvesting nodes under a power budget. Via testbed evaluation using commercial off-the-shelf energy harvesting nodes, we show that Panda outperforms existing protocols by up to 3x in terms of the neighbor discovery rate. We further explore this problem and consider a general throughput maximization problem among a set of heterogeneous energy-constrained ultra-low-power nodes. We analytically identify the theoretical fundamental limits of the rate at which data can be exchanged between these nodes, and design the distributed probabilistic protocol, EconCast, which approaches the maximum throughput in the limiting sense. Performance evaluations of EconCast using both simulations and real-world experiments show that it achieves up to an order of magnitude higher throughput than Panda and other known protocols.

We then study FD wireless - simultaneous transmission and reception at the same frequency - a key technology that can significantly improve the data rate and reduce communication latency by employing self-interference cancellation (SIC). In particular, we focus on enabling FD on small-form-factor devices leveraging the technique of frequency-domain equalization (FDE). We design, model, and optimize the FDE-based RF canceller, which can achieve >50dB RF SIC across 20MHz bandwidth, and experimentally show that our prototyped FD radios can achieve a link-level throughput gain of 1.85-1.91x. We also focus on combining FD with phased arrays, employing optimized transmit and receive beamforming, where the spatial degrees of freedom in multi-antenna systems are repurposed to achieve wideband RF SIC. Moving up in the network stack, we study heterogeneous networks with half-duplex and FD users, and develop the novel Hybrid-Greedy Maximum Scheduling (H-GMS) algorithm, which achieves throughput optimality in a distributed manner. Analytical and simulation results show that H-GMS achieves 5-10x better delay performance and improved fairness compared with state-of-the-art approaches.

Finally, we described experimentation and measurements in the city-scale COSMOS testbed being deployed in West Harlem, New York City. COSMOS' key building blocks include software-defined radios, millimeter-wave radios, a programmable optical network, and edge cloud, and their convergence will enable researchers to remotely explore emerging technologies in a real world environment. We provide a brief overview of the testbed and focus on experimentation with advanced technologies, including the integrating of open-access FD radios in the testbed and a pilot study on converged optical-wireless x-haul networking for cloud radio access networks (C-RANs). We also present an extensive 28GHz channel measurements in the testbed area, which is a representative dense urban canyon environment, and study the corresponding signal-to-noise ratio (SNR) coverage and achievable data rates. The results of this part helped drive and validate the design of the COSMOS testbed, and can inform further deployment and experimentation in the testbed.

In this thesis, we make several theoretical and experimental contributions to ultra-low-power energy harvesting networks and the IoT, and FD wireless. We also contribute to the experimentation and measurements in the COSMOS advanced wireless testbed. We believe that these contributions are essential to connect fundamental theory to practical systems, and ultimately to real-world applications, in future wireless networks.

Files

  • thumnail for Chen_columbia_0054_16253.pdf Chen_columbia_0054_16253.pdf application/pdf 8.33 MB Download File

More About This Work

Academic Units
Electrical Engineering
Thesis Advisors
Zussman, Gil
Degree
Ph.D., Columbia University
Published Here
October 20, 2020