Hi everyone, for my final project topic I’d like to look at the coverage of sensor networks, and the problem of integrating data collected from multiple local sensors.
Sensors vary in both scope and signal complexity. The information they return could be as simple as a binary flag (for example, a metal detector that beeps when a detection threshold is crossed), or as complex as a video (collected by a surveillance camera), requiring sophisticated analysis to extract information of interest.
An increasingly common application for sensors is to scan a region for a particular object or substance. One can deploy a large
number of small, coarse, “local” devices that may have uncertainties in their readings. Swarms of local sensors at micro- or nano-scale have the potential to revolutionize the way that we think about security and surveillance problems. However, this brings with it the difficulty of integration. How does one collect local information and collate it into global environmental data?
A way to think about this integration problem, in terms of topology, is what are the global features of a surface, given local data in form of a triangulation ?
Another question is, whether a certain network of sensors is able to cover the entire region of interest. That is, if the coverage of each sensor is depicted by a radially symmetric neighborhood (coverage disks), does the union of these coverage disks about the nodes cover the region of interest ? (Problem of Blanket Coverage). In this case the trade off between power usage, and the performance of the sensor network must also be considered.
jedunay
February 19, 2018 — 23:51
The topic seems interesting. When you are considering this, are you going to be thinking of the sensors as essentially being continuous (like the way we often think of water as being continuous even though it is made up of molecules) or will you still be thinking about it discretely? topology does seem like a good way to think about this.
bjiang8
February 19, 2018 — 15:54
It’s cool to use topology on data fusion. Sometimes it may be hard to find similar features in different data spaces, and we may lose structure signals by using traditional signal processing methods. Therefore, using topology seems to be a good supplement on data analysis. But it is also important to interpret features you get.