Center for Embedded Network Sensing

A broad class of applications including environmental sampling, public health monitoring, precision agriculture, and security require the ability to sense highly dynamic spatiotemporal phenomena. Solar light radiation, CO2 flux, and algal blooms are just a few examples of interesting dynamic environmental phenomena. These environmental phenomena all have high spatial and temporal variation. Currently, there are four sampling methods available for sampling dynamic spatiotemporal phenomena – static sensor sampling, deterministic actuated sensor sampling, adaptive sampling, and a combination of static sensors and actuated sensors. Spatially dynamic phenomena require an impractically large number of static sensors to be deployed which results in not only an excessive cost in resources but also has the potential to disturb the environmental phenomena under investigation. Actuated sampling methods such as a raster scan and adaptive sampling are sufficient for sampling spatially dynamic phenomena; however, this comes at the cost of increased delay (sampling latency) which makes such methods unsuitable for sampling temporally dynamic phenomena. The forth sampling method, which is a combination of static sensors and actuated sensors, is suitable for sampling dynamic spatiotemporal phenomena. Events occurring outside the range of a static sensor, however, might be missed in this approach. Thus, the performance of the system depends on the number of static sensors. Therefore, in order to accurately and efficiently characterize dynamic spatiotemporal phenomena to achieve high fidelity reconstruction, an intelligent algorithm is needed that combines resources with varying sensing and mobility capabilities.

We introduce Multitier Multiscale Sensing, which is a new paradigm for actuated sensing for efficiently sampling dynamic spatiotemporal phenomena with high ?delity. This approach introduces a hierarchy of sensors according to sampling ?delity, spatial coverage, and mobility charac¬teristics. The application of solar light radiation was chosen to illustrate how the general multitier multiscale paradigm can be implemented as a two-tier multiscale sensing technique. Experiments were performed in simulation and on a physical robotic system. Results show that multitier multiscale sensing is suitable for sampling dynamic spatiotemporal phenomena. In the future we will also apply the multitier multiscale method to a two-tier multimodal multiscale sensing approach for the study of temperature and chlorophyll in an aquatic setting. This case study uses a multimode implementation that will enable high fidelity sampling of large two- and three-dimensional environments with sparse sensing resources for dynamic field characterization. Multimode actuated sensing and multimode sensing models will be exploited to provide high fidelity reconstruction of algal bloom dynamics in aquatic environments.