Identification methods for low-power sensors

The Internet of Things (IoT) will soon result in a huge number of smart devices with actuators, sensors, network connectivity and signal processing. The scientific objective is to extend the current state-of-the-art frequency identification methods to enable their use within low-cost, low power IoT devices.

Cedric Busschotsā€™ PhD aims for a methodology to select the algorithm and implementation depending on the available program and data memory size, processing time, power and communication bandwidth of the IoT device used. This includes the selection and application of the excitation signal, the choice of the algorithm and its different possible numerical implementations. His PhD focuses first on Single Input Single Output systems, followed by extensions to multiple inputs and/or multiple output with the challenge to find the optimal balance between processing data locally, transmitting data and centralizing the final results.

Applications can be found in various engineering domains including control system design, system analysis, and fault detection. The methodology will be demonstrated on two complementary application domains, namely a lightly damped mechanical resonating system and a heavily damped medical measurement device. Both low-power/high-performance and low/high data-rate IoT systems are targeted.

References:

  1. R. Pintelon and J. Schoukens, System Identiļ¬cation: A Frequency Domain Approach, 2nd ed. John Wiley & Sons, 2012.
  2. G. H. Golub and C. F. Van Loan, Matrix Computations, 3rd ed. The Johns Hopkins University Press, 1996.
Back to top