Feature/weighted growth

examples/development.py

@@ -2,6 +2,14 @@
 
 import growbikenet as gbn
 
-edges_ranked = gbn.growbikenet("Oelde", 
-    seed_point_type='file', 
-    import_files={'street_network':"./tests/test_data/oelde_street_network.gpkg", 'seed_points':"./tests/test_data/oelde_seed_points.gpkg"})
+gbn.settings.import_data_impact = 9
+gbn.settings.import_data_trip_point_balance = 0.4 # alpha
+edges_ranked = gbn.growbikenet("Turin", 
+    export_data = True,
+    import_files={
+        'street_network':"./tests/test_data/turin_street_network.gpkg",
+        'point_data':"./tests/test_data/turin_crashes.gpkg",
+        'trip_data':"./tests/test_data/turin_trips.csv",
+        },
+    )
+print(len(edges_ranked))

growbikenet/constants.py

@@ -2,24 +2,25 @@
 
 PBI_CUSTOM_FILTER : list[str]
     Custom filter for protected bicycle infrastructure (pbi)
+EXISTING_NETWORK_MINIMUM_COMPONENT_LENGTH : int
+    Minimum length a bike network component needs to have for seed points to snap, in meters
+GRID_SPACING_TRIANGULATE : int
+    Grid spacing in meters for grid triangulation that ensures that any point in the city is always within buffer distance b=500m of the network (if seed points snap perfectly).
+GRID_SPACING_QUADRANGULATE : int
+    Grid spacing in meters for quadrangulation that ensures that any point in the city is always within buffer distance b=500m of the network (if seed points snap perfectly).
+GRID_SPACING_TRIANGLE : int
+    Grid spacing in meters for triangle grid that ensures that any point in the city is always within buffer distance b=500m of the network (if seed points snap perfectly).
+
 _PRESET_TAGS : dict
     Pre-defined tags to select tags as seed points
 _PHI_LIMITS : list[float]
     Two orientation order limits between street networks with:
     1) negligible grid elements, 2) some grid elements, 3) grid.
     We aimed to use the tercile limits from the paper [3]_ (Fig 2), but the values here are lower for unknown reasons, also with the unweighted version. Also, it was aimed to have Barcelona in the grid category. For these reasons, the limits were lowered.
-EXISTING_NETWORK_MINIMUM_COMPONENT_LENGTH : int
-    Minimum length a bike network component needs to have for seed points to snap, in meters
 _SEED_POINT_SNAP_DISTANCE_FACTOR : float
     Factor to multiply seed_point_grid_spacing with, to determine auto value of seed_point_snap_distance
 _EXISTING_NETWORK_SPACING_FACTOR : float
     Factor to multiply seed_point_grid_spacing with, to determine auto value of existing_network_spacing
-GRID_SPACING_TRIANGULATE : int
-    Grid spacing in meters for grid triangulation that ensures that any point in the city is always within buffer distance b=500m of the network (if seed points snap perfectly).
-GRID_SPACING_QUADRANGULATE : int
-    Grid spacing in meters for quadrangulation that ensures that any point in the city is always within buffer distance b=500m of the network (if seed points snap perfectly).
-GRID_SPACING_TRIANGLE : int
-    Grid spacing in meters for triangle grid that ensures that any point in the city is always within buffer distance b=500m of the network (if seed points snap perfectly).
 _BUFFER_SEED_POINTS_EXNW_FACTOR : float
     Factor to multiply existing_network_spacing with, to determine which previously determined seed points (grid or rail) to drop that are too close to the extra existing network points
 _BEARING_BINS : int
@@ -40,10 +41,16 @@
         ]
 # Populate ox.settings.useful_tags_way to make application of custom filter possible
 import osmnx as ox
-for custom_tag in ["cycleway", "bicycle", "cycleway:right", "cycleway:left", "cycleway:both", "cyclestreet"]:
+for custom_tag in ["highway", "cycleway", "bicycle", "cycleway:right", "cycleway:left", "cycleway:both", "cyclestreet"]: # This list should contain all tags used in any custom filters
     if custom_tag not in ox.settings.useful_tags_way:
         ox.settings.useful_tags_way.extend(custom_tag)
 
+EXISTING_NETWORK_MINIMUM_COMPONENT_LENGTH = 100 
+
+GRID_SPACING_TRIANGULATE = 1707 # a=2b/(2-sqrt(2))
+GRID_SPACING_QUADRANGULATE = 1000 # a=2b
+GRID_SPACING_TRIANGLE = 1154 # h/2=b=a*sqrt(3)/4 -> a=4b/sqrt(3)
+
 _PRESET_TAGS = {
             "rail": {"railway": ["station", "halt"]},
             "school": {"amenity": ["kindergarten", "school", "college", "university"]},
@@ -52,16 +59,10 @@
 
 _PHI_LIMITS = [0.02, 0.08] # Tercile limits in the paper: 0.033, 0.161
 
-EXISTING_NETWORK_MINIMUM_COMPONENT_LENGTH = 100 
-
 _SEED_POINT_SNAP_DISTANCE_FACTOR = 0.25
 
 _EXISTING_NETWORK_SPACING_FACTOR = 0.5
 
-GRID_SPACING_TRIANGULATE = 1707 # a=2b/(2-sqrt(2))
-GRID_SPACING_QUADRANGULATE = 1000 # a=2b
-GRID_SPACING_TRIANGLE = 1154 # h/2=b=a*sqrt(3)/4 -> a=4b/sqrt(3)
-
 _BUFFER_SEED_POINTS_EXNW_FACTOR = 0.5
 
-_BEARING_BINS = 72
+_BEARING_BINS = 72

growbikenet/functions.py

@@ -163,7 +163,7 @@ def _resolve_auto_parameters(
         phi,
         import_files,
     ):
-    """Resolve auto parameters and parameter inconsistencies
+    """Resolve auto parameters, their inconsistencies, and settings
 
     Parameters
     ----------
@@ -232,29 +232,35 @@ def _resolve_auto_parameters(
     if existing_network_spacing == 'auto':
         existing_network_spacing = int(np.ceil(seed_point_grid_spacing*constants._EXISTING_NETWORK_SPACING_FACTOR))
 
+    # Set import data balance to 0 or 1 if only one data set is imported
+    if import_files['point_data'] is not None and import_files['trip_data'] is None:
+        settings.import_data_trip_point_balance = 0
+    elif import_files['trip_data'] is not None and import_files['point_data'] is None:
+        settings.import_data_trip_point_balance = 1
+
     return seed_point_type, seed_point_grid_spacing, seed_point_linking, existing_network_spacing
 
 
-def add_trip_data_to_net(trips, A, crs_projected, matching_distance=500):
+def add_trip_data_to_net(trips, A, crs_projected=settings.crs_projected, matching_distance=settings.import_trip_data_snap_distance):
     """Match trip data to network edges
 
     First, match origin and destination points given in trips to the nodes. Only consider trips where both origins and nodes are matched within matching_distance. Then, for each trip, find the shortest path over the edges from matched origin node to matched destination node, and add 1 (or optionally "num" if column provided in trips) to the affected edges.
 
     Parameters
     ----------
     trips : pandas DataFrame
-        A df of unprojected origin-destination coordinates (columns: o_lat, o_lon, d_lat, d_lon), with each row encoding a trip. Optional with a column "num" containing an integer. This could be (number of) trip events. If "num" column is not provided, assumes 1 per trip.
+        A df of unprojected origin-destination coordinates (columns: o_lat, o_lon, d_lat, d_lon), with each row encoding a trip, in unprojected crs EPSG:4326. Optional with a column "num" containing an integer. This could be (number of) trip events. If "num" column is not provided, assumes 1 per trip.
     A: networkx.graph
         Graph created from triangulation edge list
     crs_projected : str
         Coordinate reference system that is used to project spatial data.
     matching_distance : int
-        Matching distance in meters
+        Matching distance in meters. Set via settings.import_trip_data_snap_distance.
 
     Returns
     -------
     graph_with_data : networkx.graph
-        The same graph created from triagulation edges list, but with a new edge attribute "num_trips" populated with the summed up "num" values of all trips where both origins and destinations could be matched to the closest network nodes within matching_distance. 
+        The same graph created from triangulation edges list, but with a new edge attribute "num_trips" populated with the summed up "num" values of all trips where both origins and destinations could be matched to the closest network nodes within matching_distance. 
     """
 
     graph_with_data = A.copy()
@@ -291,7 +297,6 @@ def add_trip_data_to_net(trips, A, crs_projected, matching_distance=500):
         if o_node == d_node or o_distance > matching_distance or d_distance > matching_distance:
             continue  
 
-
         # Get shortest path between the snapped orgin and destination nodes
         path = nx.shortest_path(graph_with_data, source=o_node, target=d_node, weight="distance")
         path_edges = list(zip(path[:-1], path[1:]))
@@ -312,7 +317,7 @@ def add_trip_data_to_net(trips, A, crs_projected, matching_distance=500):
     return graph_with_data
 
 
-def add_point_data_to_net(points, edges, crs_projected, matching_distance=500):
+def add_point_data_to_net(points, edges, crs_projected=settings.crs_projected, matching_distance=settings.import_point_data_snap_distance):
     """Match point data to network edges
 
     Parameters
@@ -321,10 +326,10 @@ def add_point_data_to_net(points, edges, crs_projected, matching_distance=500):
         A gdf of unprojected point geometries, optional having a column "num" containing an integer. This could be (number of) point events like crashes or citizen feedback to improve bike infrastructure. If "num" column is not provided, assumes 1 per point.
     edges : geopandas.geodataframe.GeoDataFrame
         A gdf of projected spatial network edges. This is the routed network of seed points.
-    matching_distance : int
-        Matching distance in meters
     crs_projected : str
         Coordinate reference system that is used to project osm data.
+    matching_distance : int
+        Matching distance in meters. Set via settings.import_point_data_snap_distance.
 
     Returns
     -------
@@ -1329,4 +1334,40 @@ def orientation_order(g_undir):
     Hg = 1.386
     Hmax = 3.584
     phi = 1 - ((Hw-Hg)/(Hmax-Hg))**2
-    return phi
+    return phi
+
+def _get_weighted_distances(B, num_types):
+    """Get weighted distances by edge attribute num_types
+
+    The calculation follows [1]_, only without +1 in the numerator and with a small epsilon to prevent division by zero.
+
+    Parameters
+    ----------
+    B : networkx.classes.multigraph.MultiGraph
+        The routed, grown bicycle network graph, where edges have the attribute num_types. The numerical attribute "distance" must exist for all edges.
+    num_types : str
+        Name of the attribute to weight the distances
+
+    Returns
+    -------
+    dist_weighted_by_types_dict : dict
+        Dictionary where keys are the edges (tuples of node ids), and values are the weighted distances following [1]_
+
+    References
+    ----------
+    .. [1] P. Folco, L. Gauvin, M. Tizzoni, M. Szell, "Data-driven micromobility network planning for demand and safety", Environment and planning B: Urban analytics and city science 50(8), 2087-2102 (2023)
+    """
+
+    num_types_dict = nx.get_edge_attributes(B, num_types)
+    dist_dict = nx.get_edge_attributes(B, "distance")
+    num_types_per_km_dict = {}
+
+    for k,d in dist_dict.items():
+        num_types_per_km_dict[k] = 1000*num_types_dict[k]/d
+    max_n = max(num_types_per_km_dict.values())+1e-10
+
+    dist_weighted_by_types_dict = {}
+    for k,d in dist_dict.items():
+        dist_weighted_by_types_dict[k] = d/(1+settings.import_data_impact*(num_types_per_km_dict[k]/max_n))
+
+    return dist_weighted_by_types_dict