Develop HiSpot open-source software to realize mathematical planning solver, approximate algorithms and heuristic algorithms to solve spatial optimization problems.
Wang, S., Zhou, J., Liang, H., Wang, Z., Su, C., & Li, X. (2023, November). A New Approach for Solving Location Routing Problems with Deep Reinforcement Learning of Emergency Medical Facility. In Proceedings of the 8th ACM SIGSPATIAL International Workshop on Security Response using GIS 2023 (pp. 50-53).https://doi.org/10.1145/3615884.3629429
LRP jointly considers the facility location problem (FLP) and the vehicle routing problem (VRP)
%% data process
import numpy as np
import random
import geopandas as gpd
region=gpd.read_file("../data/beijing/changping/changping.shp")
poi = gpd.read_file("../data/beijing/changping/changping-poi.shp")
data = poi[['lon', 'lat']]
num_rpoints = poi.shape[0]
rpoints = [(data['lon'][i], data['lat'][i]) for i in range(num_rpoints)]
rpoints_np = np.array(rpoints)
# facility
facilites = [3, 11, 27, 29, 31, 34, 40, 43, 53, 63]
rfacility_nodes_np = rpoints_np[facilites]
rfa_cap = [(random.randint(35, 40), random.randint(40, 45)) for i in range(len(facilites))]
# demand
demands = list(set(range(num_rpoints))-set(facilites))
rdemand_nodes_np = rpoints_np[demands]
rde_demand = [random.randint(1, 10) for i in range(len(demands))]
%% inference
from pulp import *
from hispot.LRP import LRP_cap
rselected, rassigned, robj = LRP_cap(facility_nodes=rfacility_nodes_np,
demand_nodes=rdemand_nodes_np,
solver=GUROBI_CMD(),
fa_cap=rfa_cap,
de_demand=rde_demand).prob_solve()
import geoplot as gplt
import geoplot.crs as gcrs
import matplotlib.pyplot as plt
%% prepare the LineString and center Points to plot the solution
from shapely.geometry import LineString
crs = 'EPSG:4326'
lines = gpd.GeoDataFrame(columns=['id', 'geometry'], crs=crs)
k = 0
for i in rassigned:
center = rfacility_nodes_np[int(i)]
for j in rassigned[i]:
assign = rdemand_nodes_np[j]
line = LineString([center, assign])
lines.loc[k] = [k+1, line]
k = k+1
centers=list(np.array(facilites)[rselected])
uncenters=list(set(facilites)-set(centers))
center_points = poi.iloc[centers]
uncenter_points = poi.iloc[uncenters]
%% plot
ax = gplt.sankey(lines,
projection=gcrs.Mollweide(),
linewidth=1,
color='green',
zorder=3,
figsize=(10, 8),)
gplt.polyplot(region,
projection=gcrs.AlbersEqualArea(),
edgecolor="white",
facecolor="#DBE4C6",
zorder=1,
ax=ax,)
gplt.pointplot(poi,
extent=region.total_bounds,
s=5,
color='#3C486B',
alpha=1,
linewidth=0,
label='POI',
zorder=2,
ax=ax)
gplt.pointplot(center_points,
extent=region.total_bounds,
s=10,
color='orange',
alpha=1,
linewidth=0,
marker='*',
label='Served Facility',
zorder=4,
ax=ax)
gplt.pointplot(uncenter_points,
extent=region.total_bounds,
s=10,
color='grey',
alpha=1,
linewidth=0,
marker='*',
label='Unserved Facility',
zorder=4,
ax=ax)
plt.legend(loc='upper left')
plt.show()
see notebook for more code details.
import random
import numpy as np
%% Generate problem with synthetic data
num_points = 10
num_hubs = 3
PC, PT, PD = 1, 1, 1
# PC, PT, PD = 1.0, 0.75, 1.25
weight = np.random.randint(1, 2, size=(num_points, num_points))
points = [(random.random(), random.random()) for i in range(num_points)]
points_np = np.array(points)
%% inference
from pulp import *
from hispot.FLP import PHub
hubs, assigns, obj = PHub(num_points=num_points,
points=points_np,
solver=PULP_CBC_CMD(),
num_located=num_hubs,
weight=weight,
collect_cost=PC,
transfer_cost=PT,
distribution_cost=PD).prob_solve()
%% plot
import matplotlib.pyplot as plt
plt.figure(figsize=(8,8))
name = 'Problem(P=' + str(num_hubs) + ',I=' + str(num_points) + ') \nThe minimum total cost =' + str(round(obj,4))
plt.title(name, fontsize = 15)
#Points
plt.scatter(*zip(*points_np), c='Blue', marker='o',s=30, label = 'Demand Points', zorder=2)
plt.scatter(*zip(*points_np[hubs]), c='Red', marker='*',s=100,label = 'Medians',zorder=3)
#Lines
for i in assigns:
center_point = points_np[i]
for j in assigns[i]:
demand_points = points_np[j]
pts = [points[i], points[j]]
plt.plot(*zip(*pts), c='Black', linewidth=2, zorder=1)
for i in hubs:
for j in hubs:
if i != j:
h = [points[i], points[j]]
plt.plot(*zip(*h), c='Lightblue', linewidth=2, zorder=1)
# plt.grid(True)
plt.legend(loc='best', fontsize = 15)
plt.show()
see notebook for plotting code and more details.
- Clone the repo
git clone https://github.com/HIGISX/hispot.git - conda create -n higis python
- conda activate higis
- Launch jupyter notebook
jupyter notebook(pip install jupyter) - pip install pulp
- pip install HiSpot-0.1.0-py3-none-any.whl
You should now be able to run the example notebooks.
You can choose to install and use another solver that is supported by Pulp:
- GLPK (included in conda environment)
- COIN-OR CBC
- CPLEX
- Gurobi
-numpy
-pulp
-higis(pip install HiSpot-0.1.0-py3-none-any)
[optional] (for plotting)
-matplotlib
-geopandas
-geoplot
pip install higis pip install numpy pip install pulp
If you are having trouble, please create an issue, start a discussion, or talk to us in the gitter room.