A Lightweight Dynamic Optimization Methodology for Embedded Wireless Sensor Networks*

This chapter proposes a lightweight dynamic optimization methodology for embedded wireless sensor networks (EWSNs), that intelligently selects appropriate initial tunable parameter value settings by evaluating application requirements, the relative importance of these requirements with respect to each other, and the magnitude in which each parameter effects each requirement. This one-shot operating state obtained from appropriate initial parameter value settings provides a high-quality operating state with minimal design space exploration for highly constrained applications. The chapter presents a dynamic optimization methodology to iteratively improve the one-shot operating state to provide an optimal or near-optimal operating state for less constrained applications. The chapter addresses a lightweight online greedy algorithm that leverages intelligent parameter arrangement to iteratively explore the design space, resulting in an operating state within 2% of the optimal operating state while exploring only 0.04% of the design space.