Fix weighing == time for OD analyses

Deltares / ra2ce

RA2CE helps to quantify resilience of critical infrastructure networks, prioritize interventions and adaptation measures and select the most appropriate action perspective to increase resilience considering future conditions.

https://deltares.github.io/ra2ce/

Other

7 stars 1 forks source link

Fix weighing == time for OD analyses #472

Closed ArdtK closed 2 weeks ago

ArdtK commented 4 weeks ago

Context During development of #433 it was found that also OD-analyses will fail when using weighing == time.

Problem If weighing == time all edges should have a valid value for attribute time. If not this message is thrown:

No path between <origin node> and <destination node> This can be triggered by calling one of these functions:

nx.shortest_path;

nx.dijkstra_path_length.

Options

Fix it in the different analyses.
Preprocess the networks so that attributes length, time and avgspeed are present on all edges. <- PREFERRED

Scope

[x] Create flow diagram on usage of avg_speed.csv
[x] Create object/reader/writer for avg_speed.csv
[x] Create methods to initiatalize this object based on a network
[ ] #485

Notes After this the code changed in #433 can be simplified further.

ArdtK commented 4 weeks ago

Can be reproduced by example_origin_destination_analysis and example_optimal_route_origin_destination:

from ra2ce.analysis.analysis_config_data.enums.weighing_enum import WeighingEnum

handler = Ra2ceHandler(network=network_ini, analysis=analysis_ini)
handler.input_config.analysis_config.config_data.analyses[0].weighing = WeighingEnum.TIME
handler.configure()
handler.run_analysis()

ArdtK commented 3 weeks ago

Analysis nx.shortest_path is used in:

origin_closest_destination (helper for multi_link_origin_closest_destination, optimal_route_origin_closest_destination)

nx.dijkstra_path_length is used in:

single_link_redundancy (#433)
multi_link_redundancy (#433)
optimal_route_origin_destination
origin_closest_destination

These work on the next inputs:

single_link_redundancy: base_graph
multi_link_redundancy: base_graph_hazard
multi_link_origin_closest_destination: origins_destinations_graph/origins_destinations_graph_hazard
optimal_route_origin_closest_destination: origins_destinations_graph_hazard
optimal_route_origin_destination: origins_destinations_graph
origin_closest_destination: origins_destinations_graph_hazard

This means that on all graphs the right attributes should be added:

base_graph
base_graph_hazard
origins_destinations_graph
origins_destinations_graph_hazard

Conclusion As the _hazard graphs are derived from the base graphs, fixing it on base_graph and origin_destinations_graph should be sufficient. Thus option 2. sounds the most viable.

ArdtK commented 3 weeks ago

Current flow avg_speed File avg_speed.csv is read/created on creation of OSM/Vector network.

Vector (vector_network_wrapper) If the network has attribute length on all edges and maxspeed on 1 or more edges:

If file avg_speed.csv exists: read it. If not: calculate it based on calculation nut.calc_avg_speed and write result to avg_speed.csv.
Apply avg_speed to the graph.

OSM (osm_network_wrapper) Same logic as for Vector.

Questions

How should this be done for other network types? (Shape, Trail)