lcpz / lfb-datasets

LFB task allocation datasets

London Fire Brigade Task Allocation Datasets

This repository contains real-world datasets that can be used to generate task allocation problems simulating the mobilisation of fire-fighting UAV swarms in the London metropolitan area.

We used the following resources:

• LFB Incident Records
• LFB Mobilisation Records
• LFB Fleet List

The data considered is dated from 1 January 2009 to 30 June 2022. The fleet list is updated to January 2022.

Station locations have been retrieved in June 2021 from Google Maps and elaborated with fastkml.

How to generate data from LFB records

• If necessary, use xlsx2csv to convert to .csv format.
• Select the following columns from the Incident Records:

  IncidentNumber,DateOfCall,TimeOfCall,IncidentGroup,Easting_rounded,Northing_rounded

• Select the following columns from the Mobilisation Records:

  IncidentNumber,(DateAnd)TimeMobilised,(DateAnd)TimeArrived,AttendanceTimeSeconds,DeployedFromStation_Code

• Exclude rows containing NULL, Null, or null.
• Intersect the datasets thus obtained on with scripts/intersect.py. Intersection order: Mobilisation, Incident.
• Convert to latitude-longitude coordinates with scripts/osgb36_to_wsg84.py. Dependency: pyproj Be sure that Easting_rounded and Northing_rounded are the last two columns in the intersection.
• If necessary, sort entries chronologically with scripts/sort.py.

How to generate problems

• Incident records: task IDs, task locations, optionally task weight (notional cost).
• Mobilisation records: task time windows (alpha: date and time arrived; gamma: attendance time in seconds; beta: value in [alpha, gamma]).
• Fleet List: agent IDs and speeds.
• Initial agent locations (station locations).

Coalition values (e.g., the utility of multiple pumping appliances working on the same fire) cannot be defined from the records.

However, we can generate them with distributions typically used in Coalition Structure Generations. For example, we can use the distributions defined in Section 4 of this paper.

Citation

If you use these datasets in your work, please use the following citation.

Luca Capezzuto (2021). "Large-scale, Dynamic and Distributed Multi-agent Coordination for
Real-time Systems". PhD Thesis. School of Electronics and Computer Science, University of
Southampton.

Licenses

OGLv2 for the records and ODBL for the fleet list.