Raw data can be shared among the partners of the project (GAMES) and must be placed in the following folders:
datasets/autoteldatasets/windkraft_simonsfelddatasets/mobility
Simulated mobility scenarios are available. For instance, the directory ./mobility/simulated contains the following files:
sim_reserv_df_in_zurich_1_day_from_userdemand.csv: Simulated reservations for one day in Zurich (generated using a pool of user demands for mobility).sim_presence_matrx_in_zurich_6_months_statistical_sampler.csv: Simulated presence matrix for six months in Zurich (generated with a statistical sampler).
- demo_traffic_model
# Defne macroscopic traffic model & street network for two geographical regions
TelAviv = Macroscopic_traffic_model(geographical_area = 'Tel Aviv, Israel')
Zurich = Macroscopic_traffic_model(geographical_area = 'Zurich, CH')
lon_min, lon_max, lat_min, lat_max = TelAviv.get_minmax_lon_lat() # get minimum and maximum latitudes, longitudes from the street graph
TelAviv.plot_graph_data(feature = 'speed_kph', data_threshold = 50) # in red are streets with speed >= 50 kph
#TelAviv.plot_graph_data(feature = 'travel_time', data_threshold = 60) # in red are streets with travel_time >= 60 seconds
#TelAviv.plot_graph_data(feature = 'length', data_threshold = 100) # in red are streets with length >= 20 meters
#TelAviv.plot_graph_data(feature = 'highway', data_threshold = 'residential') # in red are branches labelled as residential streets
df_edges = TelAviv.create_edge_dataframe() # get lat,lon start-end of the edges/streets
path = TelAviv.shortest_path_start_end(LO_LA_start= [34.79, 32.073], LO_LA_end = [34.791, 32.09] ) # get (if exist) the shortest path/route from start to end (latitude,longitude)
df_trips = pd.DataFrame([[1, 34.79, 32.073, 34.789, 32.093, 9],
[2, 34.789, 32.138, 34.7923, 32.1, 12]],
columns=['reservation_id', 'startLongitude', 'startLatitude',
'endLongitude', 'endLatitude', 'distance']) # example data set with two trips with only (ID,lat,lon,dis)
TelAviv.plot_trip_routes(df_trips, show=True, route_alpha=0.9)
routes, df_route_features = TelAviv.get_shortest_routes_and_features(df_trips) # routes= list of routes, df_route_features= data frame with features of the trip and shortest routes- demo_data_loader_pre_process
df_autotel = data_loader(data_dir='datasets/autotel', file_name='autotel_2021_2022.pkl')
df, df_sequence = preprocess_trip_data_frame(df,TelAviv.get_minmax_lon_lat())
results_daily = get_daily_profiles_data(df_sequence)
matrix_day = results_daily['matrix_daily_departures'] # a [n_days x 24] array contining the total number of departures for each day and hour in the data set 
- demo_zoning_analysis
# example of macroscopic zoning
n_disretized = [15, 15]
# grid partition of lat and lon.... n_disretized[0] x n_disretized[1] ....append zone lat,lon and indices to the dataframe
df_sequence_with_zones, linspace_lat, linspace_lon = define_zones(df_sequence, n_disretized_lat_lon=n_disretized)
# get statistics of idle times and trip durations on each zone
Idle_duration_zone_stats, Trips_duration_zone_stats = matrix_stats_idle_duration(df_sequence_with_zones)
# plot map (parking times)
plot_zone_duration_stats(Idle_duration_zone_stats['mean'], Idle_duration_zone_stats['std'], TelAviv, label1='Mean idle time', label2='STD idle time')
# plot inflows and outflows
plot_density_arrivals_departures_net_out_flows(Trips_duration_zone_stats['n_samples'],
Idle_duration_zone_stats['n_samples'], TelAviv)
- demo_train_total_mobility_demand_forecaster
Predict_net_number_of_departures.mp4
- demo_space_time_probabilistic_forecaster
- demo_max_coverage for EVs/Stations allocation
Gtr = Macroscopic_traffic_model(geographical_area='Tel Aviv, Israel')
data = np.array(df[['LONs','LATs']].values)
points = data[np.random.randint(0, len(data), (1,1000))]
# Number of sites to select
K, M = 80, 500
# Service radius of each site
radius = 0.0035
# Candidate site size (random sites generated)
# Run mclp opt_sites is the location of optimal sites and f is the points covered
opt_sites, f = mclp(points, K, radius, M)
fig, ax = ox.plot_graph(Gtr.G_drive, node_alpha=0.1, bgcolor="#808080", node_color='k', edge_color='k', show=False,
edge_alpha=0.3)
ax.scatter(points[:, 0], points[:, 1], 20, marker='x', c='b', alpha=0.9)
plt.scatter(opt_sites[:, 0], opt_sites[:, 1], 10, c='r', marker='+')
for site in opt_sites:
circle = plt.Circle(site, radius, color='r', fill=True, lw=2, alpha=0.3)
ax.add_artist(circle)
ax.axis('equal')
ax.tick_params(axis='both', left=False, top=False, right=False,
bottom=False, labelleft=False, labeltop=False,
labelright=False, labelbottom=False)
plt.show()
def allocate_stations(start_coords,
end_coords,
n_stations=10,
method='voronoi_k_mean',
visualize=False):
""" devides the area in clusters and voronoi partitions
it uses - kmean clustering of departures to define 'good' coordinates for the stations voronoi-like partitioning uses kdtree to assign to end-dooeds a cluster ids"""
centers, node_start, node_end, closest_dist_station = [], [], [], []
if method == 'voronoi_k_mean':
centers = kmeans(start_coords, n_stations)[0] # K-means clustering
node_start = vq(start_coords, centers)[0]
vor = Voronoi(centers)
voronoi_kdtree = cKDTree(centers) # closest distance lookup
closest_dist_station, node_end = voronoi_kdtree.query(end_coords)
if visualize:
# Plotting
fig, ax = plt.subplots()
# fig, ax = ox.plot_graph(Gtr.G_drive, node_alpha=0.1, bgcolor="#cccccc", node_color='k', edge_color='k', show=False, edge_alpha=0.1)
plot_data_points(start_coords, ax=ax)
voronoi_plot_2d(vor, ax=ax, show_vertices=False, show_points=False)
ax.tick_params(axis='both', left=False, top=False, right=False,
bottom=False, labelleft=False, labeltop=False,
labelright=False, labelbottom=False)
[plt.scatter(start_coords[node_start==cl][:,0], start_coords[node_start==cl][:,1]) for cl in np.unique(node_start)]
plot_kmeans_clustering(centers, ax=ax, alpha=0.99)
plt.show()
return centers, node_start, node_end, closest_dist_station
