Test data for spatial field reconstruction

Reiner Jedermann, University of Bremen, Germany, October 2016

Wireless sensor networks (WSN) are often applied to estimate the distribution of a spatial field, but they provide only measurements at a limited set of points. The task of spatial field reconstruction includes smoothing and interpolation of the data for any point in space. For compressed data transmission it is necessary to describe the field by a reduced set of parameters. Two example algorithms have been described in (Jedermann 2016). Test data for comparison of the accuracy of the different algorithms were generated by COMSOL simulation.

The following 2-dimensional test data set contains a simulation of 3 heat sources in a water basin, including the effects of diffusion, horizontal heat transfer by a small flow (advection) and the influence of walls. The data set can be used for own studies under the condition that the original article (Jedermann, 2016) is quoted. Feedback on improved algorithms is encouraged. The data can be either exported from the COMSOL simulation or the generated MATAB files can be directly used.

Before testing algorithms for spatial field reconstruction, a limited set of sensors has to be selected from the data (e.g. 200) and noise has to be added (e.g. sigma=0.1). We recommend using the MATLAB script ‘loadTestData.m’ for this purpose. The script can either generate a new noise instance and new sensor positions for each call (mode ‘p’), or repeat always the same values (mode ‘r’). The first 4 sensors were fixed to the corners of the field in the latter case; all other sensor locations were set randomly.

The clean reference data without noise should only be used to evaluate the accuracy of the estimated field.

Remarks on COMSOL files

The data set was generated with COMSOL version 5.0. The data set used in the article was generated with ‘fine’ mesh size with the variable ‘size1’ reduced from 0.099 to 0.025.

Advection in fluids is present in most realistic scenarios and should be included in the simulation. Nevertheless, it can be switched off by setting the Global-->Definitions-->Parameters ‘FlowOn’ to zero.


See [pdf] for more details.


         [Comsol File]

         [Matlab Files (Zip)]

         [Large set 256 by 256]

Please quote the following article, if you use this data set:

Jedermann, R.; Paul, H.; Lang, W.: Compressed radio transmission of spatial field measurements by virtual sensors. In: 2016 IEEE International Conference on Wireless for Space and Extreme Environments (WiSEE), Aachen, Germany, 2016, pp. 184-189.  (doi: 10.1109/WiSEE.2016.7877329).