2B.1 Simulation and Data Analytics with application to MoPED and other in situ sensing paradigms

Tuesday, 24 January 2017: 8:30 AM
611 (Washington State Convention Center )
Dmitry Tislin, GST, Inc., Greenbelt, MD; and S. Chettri and J. D. Evans
Manuscript (554.4 kB)

This abstract discusses the theory and algorithmic considerations for the simulation of Mobile Platform Environmental Data (MoPED) and its extensions to non-vehicular data. It extends the argument to data analytical methods enabled by the simulated data.

The MoPED system is a vehicle-based mobile platform environmental observation network operating in the US. Currently a commercial fleet of trucks, about 600 at last count, using sensor packages and communications equipment, provides this environmental data to the MoPED system. These commercial vehicles travel interstate, state and local routes and also have a footprint in metropolitan areas.

The MoPED simulator creates a fleet of anywhere between 1000-10,000 vehicles that start and finish at random times and move at random speeds along a US highway network extracted from a geospatial database of our design. Instrument packages on these moving platforms sample surface temperatures and pressures from an environmental field produced by NOAAs High-Resolution Rapid Refresh (HRRR), an hourly updated model with a 3-km resolution. Our analysis of MoPED data shows that sensors typically have noise characteristics that lead to drop-outs, drift and physically unrealizable values. Accordingly, our simulation produces temperature and pressure values that are corrupted with these kinds of noise (Figure 1). Furthermore, the simulator obtains Automated Surface Observing System (ASOS) collected data sets for data analytical purposes as described below.

The simulator permits production of MoPED-like environmental data stream (surface temperature and pressure to start with) enabling studies of the following:

1)      Data analytical methodology, including comparing noisy truck data with ASOS data for the purposes of calibration that is coincident temporally and in spatial proximity, 10-km in our studies (Figure 2).

2)      Following item 1), calibration between truck sensors when they are near-coincident spatially.

3)      Detection of spatial structure at the surface such as dry lines; for example, with the addition of a relative humidity sensor to the instrument package, one might ask the question: How many trucks might one need to obtain the structure of a dry line.

4)      The data management infrastructure for MoPED, including geospatial databases, scaling with number of trucks and data structures needed to keep track of the calibration state of sensors on trucks.

5)      Expansion to mobile sensors not on trucks, e.g., human carried cell phones.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner