| Summary: | A sensor response-based localization technology that uses the non-geotagged responses of environmental object detection sensors to estimate the positions of sensors in a defined observation area was proposed in a previous study. That method is based on the concept that the proximity relationships among sensors obtained from the closeness of the sensor response times or collected data similarities. However, since that study was limited to simulation-based experiments, this study extends the scheme to experiments involving real devices, simulated environments, and ideal environments to ascertain the accuracy of localization estimates produced using real devices. We also aimed at extracting the problems that must be resolved before the proposed technology can be put into practical use. Even if clock synchronization between the sensors was impossible, the proximity relationships among them could be estimated using before-after detection time relationships, thus confirming that sensor locations could be estimated based on these proximity relationships. In a real device experiment using 20 sensors indoors at 20.25 <inline-formula> <tex-math notation="LaTeX">$m^{2}$ </tex-math></inline-formula>, we confirmed that the average localization error was 0.84 m. Based on those results and other findings resulting from this study, we determined that the primary factor related to increased sensor localization errors was the number of incorrectly constructed sensor pairs. The results of this study are expected to facilitate the realization of physical world sensing using sensor response-based localization technology.
|
|---|