A One-Shot Dynamic Optimization Methodology and Application Metrics Estimation Model for Wireless Sensor Networks

Wireless sensor networks (WSNs), consisting of autonomous sensor nodes, have emerged as ubiquitous networks that span diverse application domains (e.g., health care, logistics, defense) each with varying application requirements (e.g., lifetime, throughput, reliability). Typically, sensor-based platforms possess tunable parameters (e.g., processor voltage, processor frequency, sensing frequency), which enable platform specialization for particular application requirements. WSN application design can be daunting for application developers, which are oftentimes not trained engineers (e.g., biologists, agriculturists) who wish to utilize the sensor-based systems within their given domain. Dynamic optimizations enable sensor-based platforms to tune parameters in-situ to automatically determine an optimized operating state. However, rapidly changing application behavior and environmental stimuli necessitate a lightweight and highly responsive dynamic optimization methodology. In this paper, we propose a very lightweight dynamic optimization methodology that determines initial tunable parameter settings to give a high-quality operating state in one-shot for timecritical and highly constrained applications. We compare our one-shot dynamic optimization methodology with other lightweight dynamic optimization methodologies (i.e., greedyand simulated annealing-based) to provide insights into the solution quality and resource requirements of our methodology. Results reveal that the one-shot solution is within 8% of the optimal solution on average. To assist dynamic optimizations in determining an operating state, we propose an application metric estimation model to establish a relationship between application metrics (e.g., lifetime, throughput) and sensor-based platform parameters. Keywords-Wireless sensor networks, dynamic optimization, application metrics estimation