2019-04-16

IRP Update

Played with OMNeT++ and SUMO a bit. Getting my bearings on how to put
together a complete simulation using these tools.

I think the next step will be to write a script to convert APRS data
into a list of fixed station locations and O-D matrices for the mobile
stations.

In short, I need to develop an algorithm for detecting vehicle "stops"
from low resolution GPS tracking data and use those to generate trips.

Those trips can then be fed into SUMO to simulate route selection,
potentially even in the presence of other traffic conditions.

SUMO is then able to simulate demand for edges within the road
network, and place simulated vehicles at locations along those routes
during the course of a simulation. It appears that Veins is able to
then interact with these simulations via the TraCl interface (via a
Python library which wraps a TCP API) to convert the simulated vehicle
locations to radio connectivity estimations.

I plan to start out by running 802.11p simulations against my
APRS-derived traffic simulations mainly because the 802.11p models are
already available, and have seen relatively recent development
attention. Then, I'll likely branch out into exploring my own protocol
stack, maybe beginning with adding a Store-Carry-Forward (SCF)
protocol layered above 802.11p. It appears that there may be some DTN
implementations available to use already as well (DTN2 and
IBR-DTN). Eventually, I may work may way back down to exploring the
PHY layer more directly with some of the modems available in Fldigi.


A note on privacy:

Given the sensitivity of this derived O-D data, I will likely not
share it directly as an output of this research.

It looks like there has been some work done on aggregating GPS data
and deriving OD matrix data in a way that preserves k-anonymity, but I
think the required level of effort needed to construct such a pipeline
may push it beyond the scope of this project.

Instead, I will likely use the APRS-derived O-D matrix data to inform
some fully artificial simulations with (*waves hands*) similar
properties. For more formal research, an anonymization technique or a more
formal description of the relevant properties of the data may be
preferable. Perhaps those amount to the same thing.