PhD Candidate: Md Kowsar Hossain
Abstract
Wireless networks form the backbone of emerging cyber–physical systems (CPS), yet ensuring both real-time guarantees and high reliability under the challenges of probabilistic link losses and complete link failures remains an open problem. This thesis introduces node programs that enable wireless networks to tolerate transient packet losses and complete link failures, ensuring reliable, real-time communication in CPS. Chapters 2-5 focus on handling transient losses, while Chapter 6 extends the approach to address complete link failures.
First, we introduce node communication programs as a replacement for traditional fixed schedules. These programs combine slot reservation with local adaptation. Next, we develop Recorp, a receiver-oriented node program framework that enables nodes to locally reallocate the retransmissions of flows in response to variations in link quality, increasing real-time capacity 50% to 140% and decreasing latency 27% to 70% versus static schedules. Building on these results, WARP enables on-the-fly synthesis of identical node programs in milliseconds, yielding up to 48% throughput gains over Recorp and reducing the time to add new flows by 5 times over a state-of-the-art centralized control plane. Next, we introduce Batching, a novel program-based approach that provides a better interface between the probabilistic behavior of the networks and the discrete nature of the scheduling, achieving up to 3 times higher throughput than single packet approach under lossy conditions. The last chapter (Fallback approach) extends our program-based scheduler to handle complete link failures, demonstrating quick recovery from up to K arbitrary failures.
Extensive simulation and testbed evaluations validate that our Recorp, WARP, and Batching approaches meet reliability and deadline targets under transient packet losses, significantly improving network latency and capacity, and our Fallback approach provides robust recovery from complete link failures.
Advisor: Octav Chipara