Fast Design Space Exploration of Vibration-Based Energy Harvesting Wireless Sensors

An energy-harvester-powered wireless sensor node is a complicated system with many design parameters. To investigate the various trade-offs among these parameters, it is desirable to explore the multi-dimensional design space quickly. Because of the large number of parameters and costly simulation CPU times, it is, however, often difficult or even impossible to explore the design space via simulation. This paper presents a response surface model (RSM) based technique for fast design space exploration of a complete wireless sensor node powered by a tunable energy harvester. As a proof of concept, a software toolkit is developed that implements the proposed design flow and incorporates either real data or parametrized models of the vibration source, the energy harvester, tuning controller, and wireless sensor node. Several test scenarios are considered, which illustrate how the proposed approach permits the designer to adjust a wide range of system parameters and evaluate the effect almost instantly but still with high accuracy. In the developed toolkit, the estimated CPU time of one RSM estimation is 25 μs and the average RSM estimation error is less than 16.5%.