•   When: Thursday, December 13, 2018 from 11:00 AM to 01:00 PM
•   Speakers: Arda Gumusalan
•   Location: ENGR 4201
•   Export to iCal

Real-time low-power wireless monitoring are increasingly being used in applications such as: Industrial Internet-of-Things, Smart City technologies, and critical infrastructure monitoring. Creating a deadline driven scheduling while considering energy management creates complex optimization problems. My research integrates energy saving mechanisms with real-time scheduling for time critical WSNs to save energy while meeting the desired quality of service requirements. I investigate and improve the energy consumption of real-time wireless sensor network (WSN) protocols utilized in industrial control systems. My contributions are in three distinct areas.

First, I focus on single cluster real-time WSNs specifically on improving time-slotted superframe based techniques. The current wireless standards for Industrial Control Networks (ICNs) employ static slots for their superframe structure. I study the concept of dynamic readjustment of time-slots to minimize the overall energy consumption and combine real-time performance with novel energy conservation methods by describing a set of dynamic modulation scaling (DMS) based adaptive packet transmission scheduling algorithms that reclaim unused slot times. To support my reclaiming method in a wireless environment, I introduce a novel low-power listening technique called Hybrid Low-Power Listening (HLPL) protocol. I evaluate my algorithms using Castalia simulator against an oracle-based approach, and show that my dynamic slot reclaiming approach, coupled with HLPL, can introduce substantial power savings without sacrificing real-time support. In order to further expand applicability scope of my solution, my dissertation work considers non-deterministic workloads.

Next, I take a deeper look into DMS and its effect on low power wireless communication. In recent years a number of studies have suggested that DMS techniques can reduce energy consumption in low-power wireless transmission technologies. These studies tend to rely on theoretical or simulation DMS models to predict network performance metrics.  However, there is little, if any, work that is based upon empirically verified network performance outcomes using DMS.  My dissertation fills that gap. First, by using GNU-Radio and SDR hardware I show how to emulate DMS in low power wireless systems.  Second, I measure the impact of varying Signal-to-Noise levels on throughput and delivery rates for different DMS control strategies.  Third, using DMS I quantify the impact of distance and finally, I measure the impact of different elevations between sender and receiver on network performance.  My results provide an empirical basis for future work in this area.

For the third part, I investigate transmission scheduling of multi-hop time critical WSNs. Previous work has shown that connection driven topology control has tremendous potential to decrease energy consumption and/or latency. DMS changes transmission energy levels and has a direct impact on packet loss rate and propagation distance. However, current work does not provide any multi-cluster communication solutions which incorporate DMS into already managed transmission energy level control. I address this gap by first formulating Mixed Integer Nonlinear Formulation (MINLP) of DMS enabled transmission scheduling for deadline driven networks. Next, I present two polynomial time heuristics. I compare them against the optimal solution by integrating the empirical measurements obtained from my SDR tests and present how DMS can be applied to multi-hop WSNs to save substantial amount of energy.