Emerging wireless networks are envisioned to simultaneously support both delay-tolerant and delay-sensitive applications. Unfortunately, providing delay guarantees in wireless networks has been a challenging problem due to (i) the unreliable and resource-constrained wireless channel, and (ii) the inherent complex interactions across multiple protocol layers. The goal of this CAREER project is to provide an advanced suite of theoretical tools, algorithms, and protocols for supporting delay-sensitive applications on wireless networks. By combining large-deviation with Lyapunov stability, this project first develops a new unified theory for characterizing the delay performance of advanced MAC/PHY mechanisms. This is then integrated into a unified cross-layer control and optimization framework for delay-sensitive applications, which leads to high-performance algorithms and protocols for MAC scheduling, advanced rate adaptation, admission control, multi-path routing, and pricing-based control. Finally, in one of the focused application areas, vehicular networks, the project develops tailored analytical tools and control algorithms to address the unique real-time requirements of highway safety applications.
Research results will have a broad impact on the wireless industry by providing novel analytical techniques and practical control protocols for wireless networks supporting delay-sensitive applications. Through the Center for Wireless Systems and Applications, the PI will actively share the results with industrial collaborators. The research is tightly integrated into the undergraduate and graduate curricula at Purdue. In particular, a vertically integrated Project team is formed to involve undergraduate students in research. The project will also benefit our society by improving highway safety and saving human lives.
Innovative and exciting applications of wireless networking are penetrating every aspect of our lives, changing the way we talk, work, play, and commute. Many of these emerging applications require stringent delay guarantees to achieve the desired performance (e.g., wireless video communications and/or safety-message exchange in vehicular networks). Unfortunately, providing delay guarantees in wireless networks has been a challenging problem due to (i) the unreliable and resource-constrained wireless channel, and (ii) the inherent complex interactions across multiple protocol layers. In this NSF CAREER project, the project team developed an advanced suite of theoretical tools, algorithms, and protocols for supporting delay-sensitive applications on wireless networks. Starting from the bottom MAC/PHY layer through the entire protocol stack, this project built a theoretical foundation for understanding delay performance, and for designing efficient and low-delay algorithms/protocols tailored to application-specific needs. Intellectual Merits: The research activities of the project have been carried on along three complementing threads. 1. The project team developed innovative methods to understand the delay-performance of the wireless control algorithms that have been proposed in the literature. Note that many of these control algorithms have been designed to maximize only the throughput of the system, without considering the delay performance. As a result, their performance under stringent delay requirements is often not well-understood and difficult to predict. To address this open problem, the project team developed a new unified theory combining large deviations with Lyapunov functions for characterizing the delay performance in complex wireless networks. This approach can be applied to both cellular networks and ad hoc wireless networks. Thus, it provides a solid theoretical foundation for understanding and improving the delay performance of wireless systems. 2. Guided by the new understanding of wireless delay performance, the project team developed new control algorithms that can achieve superior delay performance compared to the state-of-the art. A particular focus is on applications with deadline constraints, e.g., in video conferencing, each video frame must be received by the intended receiver within a bounded time. The project team designed algorithms that can achieve the highest system throughput under deadline constraints. The team first studied a single cell with a single wireless channel, and developed the first result in the literature to precisely characterize the delay-optimal wireless scheduling algorithms. Then, new control algorithms with even better performance were developed for cellular networks with multiple channels, for multi-hop wireless networks with a tree topology, and for vehicular networks with delay-tolerant traffic. These algorithms greatly improve the capacity and delay performance of next-generation wireless networks and provide better quality-of-experience for applications with differing delay requirements. 3. Throughout the above work, a challenging tradeoff is observed between complexity, throughput and delay in wireless algorithm design. While all three performance dimensions are important for practical applications, no existing algorithms can achieve high throughput, low delay and low complexity all at the same time. Thus, these existing solutions cannot meet the high performance requirement in large-scale wireless networks. To address this open challenge, the project team designed a new class of low-complexity control algorithms that can simultaneously achieve high throughput and low delay. This new class of algorithms exploits the multiple channels in modern wireless systems to achieve dramatically improved performance, while at the same time they are easy-to-implement thanks to the low overhead. Broader Impact: The techniques developed in this project, and the theory underlying them, will have a broad impact on the wireless industry, and will also benefit our society by enabling innovative delay-sensitive wireless applications. The results were disseminated through publications, seminars at interference conferences and industry partners, and graduate course materials. Four Ph.D. students have been supported by the project and were trained on both the theoretical analysis and practical algorithm design. Some of the results developed have been incorporated into graduate-level networking course at Purdue. Undergraduate students participated in the project through a VIP (Vertically Integrated Projects) course at Purdue, where they worked with graduate students as a team to experiment with wireless video delivery.
