## New density output processor

I plan to design/implement a new density processor. The difference to the existing ones is

- generalize observable area to polygon shapes
- measure at multiple fixed points inside the polygon. Current processors i) average density over a specified area (i.e. scalar) or ii) compute the density for every pedestrian in every timestep. Because of point 1) I want to use an unstructured grid; the prefered algorithm is the already developed & implemented @BZoennchen (has to be merged before)

Something like this, where each vertex in the triangulation is a point to evaluate the density:

The output would be something like:

timeStep | point ID | density |
---|---|---|

1 | 1 | 0.5 |

1 | 2 | 0.6 |

1 | 3 | 0.1 |

2 | 1 | 0.9 |

2 | 2 | 0.6 |

2 | 3 | 0.8 |

##### Potential options to integrate into the processor (only implemented when needed):

- true/false to consider agents inside or outside the defined polygon
- set density algorithm (use available ones)
- options to forward to the triangulation algorithm (be able to give the accuracy of the density sampling)
- options to forward to the density algorithm (e.g. cut-off for Gaussian density)
- only compute the density of every
`X`

timestep (as this can be quite an expensive operation) - allow averaging the density to one point (maybe this would be another processor)

##### Other issues:

- To reconstruct the density of a polygon, it is desirable to write a description of either the points or the entire triangulation in either the data file or in a separate file.

##### Aim:

Have a density evolution over time at fixed points in the scenario. I plan to investigate how many dimensions are needed to represent the density (use PCA) and use the output for QoI.