In [1]:
Copied!
"""OpenStreetMap → Sionna RT → DeepMIMO: No-Blender Urban Pipeline."""
"""OpenStreetMap → Sionna RT → DeepMIMO: No-Blender Urban Pipeline."""
Out[1]:
'OpenStreetMap → Sionna RT → DeepMIMO: No-Blender Urban Pipeline.'
OpenStreetMap → Sionna RT → DeepMIMO: No-Blender Urban Pipeline¶
What this notebook covers:
- Define a GPS bounding box around any urban area
- Download building footprints from OpenStreetMap — no Blender required
- Visualize the scene as a 2D footprint map and in Sionna RT's 3D renderer
- Run Sionna RT 2.0 ray tracing on that scene
- Export and convert the results to a DeepMIMO dataset
- Visualize the reconstructed DeepMIMO scene and channel data
Why no Blender?
The traditional pipeline uses Blender + the blosm addon to convert OSM data
into a Mitsuba XML scene. This notebook skips Blender entirely:
deepmimo.pipelines.osm_to_mitsuba.generate_scene queries the Overpass API
directly, extrudes building polygons with NumPy, and writes a valid Mitsuba
scene file that Sionna RT can load immediately.
Requirements:
pip install 'deepmimo[sionna]'
In [2]:
Copied!
%pip install 'deepmimo[sionna]' # uncomment if not installed
%pip install 'deepmimo[sionna]' # uncomment if not installed
/home/joao/DeepMIMO/.venv/bin/python: No module named pip
Note: you may need to restart the kernel to use updated packages.
Imports¶
In [3]:
Copied!
from __future__ import annotations
from pathlib import Path
import matplotlib.pyplot as plt
import numpy as np
import sionna.rt as sionna_rt
from sionna.rt import Camera, PathSolver, PlanarArray, Receiver, Transmitter
import deepmimo as dm
from deepmimo.exporters.sionna_exporter import sionna_exporter
from deepmimo.pipelines.osm_to_mitsuba import generate_scene
from deepmimo.pipelines.txrx_placement import gen_plane_grid
from deepmimo.pipelines.utils.geo_utils import convert_GpsBBox2CartesianBBox
from deepmimo.pipelines.utils.pipeline_utils import get_origin_coords
from __future__ import annotations
from pathlib import Path
import matplotlib.pyplot as plt
import numpy as np
import sionna.rt as sionna_rt
from sionna.rt import Camera, PathSolver, PlanarArray, Receiver, Transmitter
import deepmimo as dm
from deepmimo.exporters.sionna_exporter import sionna_exporter
from deepmimo.pipelines.osm_to_mitsuba import generate_scene
from deepmimo.pipelines.txrx_placement import gen_plane_grid
from deepmimo.pipelines.utils.geo_utils import convert_GpsBBox2CartesianBBox
from deepmimo.pipelines.utils.pipeline_utils import get_origin_coords
Configuration¶
Pick any GPS bounding box. A block of ~300-600 m on a side works well - large enough to capture reflections, small enough to run quickly.
The example below covers part of Munich city center, Germany.
To use a different location, replace the BBOX coordinates with any
GPS bounding box of your choice.
In [4]:
Copied!
# GPS bounding box — ~400 m x ~400 m block in Munich city center
BBOX = {
"minlat": 48.1355,
"minlon": 11.5735,
"maxlat": 48.1395,
"maxlon": 11.5795,
}
CARRIER_FREQ = 3.5e9 # 3.5 GHz
# Sionna RT ray-tracing settings
MAX_DEPTH = 2
N_SAMPLES = 1_000_000
# Transmitter: center of the area at 25 m height
# (0, 0) is the local coordinate origin = center of the GPS bbox
TX_POS = np.array([[0.0, 0.0, 25.0]])
# UE grid settings
UE_HEIGHT = 1.5 # metres above ground
GRID_SPACING = 30.0 # metres between UE positions
# GPS bounding box — ~400 m x ~400 m block in Munich city center
BBOX = {
"minlat": 48.1355,
"minlon": 11.5735,
"maxlat": 48.1395,
"maxlon": 11.5795,
}
CARRIER_FREQ = 3.5e9 # 3.5 GHz
# Sionna RT ray-tracing settings
MAX_DEPTH = 2
N_SAMPLES = 1_000_000
# Transmitter: center of the area at 25 m height
# (0, 0) is the local coordinate origin = center of the GPS bbox
TX_POS = np.array([[0.0, 0.0, 25.0]])
# UE grid settings
UE_HEIGHT = 1.5 # metres above ground
GRID_SPACING = 30.0 # metres between UE positions
Step 1 — Generate the Mitsuba Scene from OSM¶
generate_scene queries the Overpass API for building footprints within the
GPS bounding box, extrudes each polygon into a 3D PLY mesh, creates a flat
ground plane, and writes a scene.xml that Sionna RT can load directly.
An osm_gps_origin.txt file records the local coordinate origin (center of
the bbox) so that GPS coordinates can be recovered later.
In [5]:
Copied!
# Use a persistent folder so the Overpass download is cached between runs.
# Change WORK_DIR to any writable path on your system.
WORK_DIR = Path.home() / ".cache" / "deepmimo" / "osm_pipeline_munich"
WORK_DIR.mkdir(parents=True, exist_ok=True)
scene_folder = str(WORK_DIR)
osm_scene_folder = generate_scene(
minlat=BBOX["minlat"],
minlon=BBOX["minlon"],
maxlat=BBOX["maxlat"],
maxlon=BBOX["maxlon"],
scene_folder=str(WORK_DIR / "osm_scene"),
verbose=True,
)
print(f"\nScene folder : {osm_scene_folder}")
n_buildings = sum(1 for f in Path(osm_scene_folder, "meshes").glob("building_*.ply"))
print(f"Buildings written: {n_buildings}")
# Use a persistent folder so the Overpass download is cached between runs.
# Change WORK_DIR to any writable path on your system.
WORK_DIR = Path.home() / ".cache" / "deepmimo" / "osm_pipeline_munich"
WORK_DIR.mkdir(parents=True, exist_ok=True)
scene_folder = str(WORK_DIR)
osm_scene_folder = generate_scene(
minlat=BBOX["minlat"],
minlon=BBOX["minlon"],
maxlat=BBOX["maxlat"],
maxlon=BBOX["maxlon"],
scene_folder=str(WORK_DIR / "osm_scene"),
verbose=True,
)
print(f"\nScene folder : {osm_scene_folder}")
n_buildings = sum(1 for f in Path(osm_scene_folder, "meshes").glob("building_*.ply"))
print(f"Buildings written: {n_buildings}")
Scene already exists at /home/joao/.cache/deepmimo/osm_pipeline_munich/osm_scene, skipping generation. Scene folder : /home/joao/.cache/deepmimo/osm_pipeline_munich/osm_scene Buildings written: 265
Visualize — 2D Building Footprint Map¶
Before loading the scene into Sionna we can plot the building footprints directly from the PLY meshes. The bottom-ring vertices (z ≈ 0) of each mesh form the building outline on the ground plane.
Google satellite view: if you have a Google Maps Static API key, you can overlay a satellite image with
deepmimo.pipelines.utils.geo_utils.fetch_satellite_view(...).
In [6]:
Copied!
meshes_folder = Path(osm_scene_folder) / "meshes"
fig, ax = plt.subplots(figsize=(8, 8))
for ply_file in sorted(meshes_folder.glob("building_*.ply")):
lines = ply_file.read_text(encoding="utf-8").splitlines()
hdr_end = next(i for i, ln in enumerate(lines) if ln == "end_header")
n_verts = int(next(ln.split()[-1] for ln in lines if "element vertex" in ln))
# Bottom-ring vertices are the first half of the vertex list (z = 0)
z_tol = 0.01 # ground-level tolerance in metres
min_poly_pts = 3
footprint = [
tuple(map(float, lines[hdr_end + 1 + i].split()))
for i in range(n_verts)
if abs(float(lines[hdr_end + 1 + i].split()[2])) < z_tol
]
if len(footprint) >= min_poly_pts:
xs = [v[0] for v in footprint] + [footprint[0][0]]
ys = [v[1] for v in footprint] + [footprint[0][1]]
ax.fill(xs, ys, alpha=0.5, color="steelblue", edgecolor="navy", linewidth=0.4)
ax.scatter(*TX_POS[0, :2], c="red", marker="^", s=250, zorder=5, label="TX")
ax.set_xlabel("x (m)")
ax.set_ylabel("y (m)")
ax.set_title("OSM Building Footprints — Munich City Center")
ax.set_aspect("equal")
ax.grid(visible=True, alpha=0.3)
ax.legend()
plt.tight_layout()
plt.show()
meshes_folder = Path(osm_scene_folder) / "meshes"
fig, ax = plt.subplots(figsize=(8, 8))
for ply_file in sorted(meshes_folder.glob("building_*.ply")):
lines = ply_file.read_text(encoding="utf-8").splitlines()
hdr_end = next(i for i, ln in enumerate(lines) if ln == "end_header")
n_verts = int(next(ln.split()[-1] for ln in lines if "element vertex" in ln))
# Bottom-ring vertices are the first half of the vertex list (z = 0)
z_tol = 0.01 # ground-level tolerance in metres
min_poly_pts = 3
footprint = [
tuple(map(float, lines[hdr_end + 1 + i].split()))
for i in range(n_verts)
if abs(float(lines[hdr_end + 1 + i].split()[2])) < z_tol
]
if len(footprint) >= min_poly_pts:
xs = [v[0] for v in footprint] + [footprint[0][0]]
ys = [v[1] for v in footprint] + [footprint[0][1]]
ax.fill(xs, ys, alpha=0.5, color="steelblue", edgecolor="navy", linewidth=0.4)
ax.scatter(*TX_POS[0, :2], c="red", marker="^", s=250, zorder=5, label="TX")
ax.set_xlabel("x (m)")
ax.set_ylabel("y (m)")
ax.set_title("OSM Building Footprints — Munich City Center")
ax.set_aspect("equal")
ax.grid(visible=True, alpha=0.3)
ax.legend()
plt.tight_layout()
plt.show()
Step 2 — Load the Scene in Sionna RT¶
In [7]:
Copied!
xml_path = str(Path(osm_scene_folder) / "scene.xml")
scene = sionna_rt.load_scene(xml_path, merge_shapes=False)
scene.frequency = CARRIER_FREQ
print(f"Scene loaded: {len(scene.objects)} objects")
print(f"Sample materials: {[o.radio_material.name for o in list(scene.objects.values())[:4]]}")
# Single isotropic antenna for both TX and RX
single_ant = PlanarArray(
num_rows=1,
num_cols=1,
vertical_spacing=0.5,
horizontal_spacing=0.5,
pattern="iso",
polarization="V",
)
scene.tx_array = single_ant
scene.rx_array = single_ant
xml_path = str(Path(osm_scene_folder) / "scene.xml")
scene = sionna_rt.load_scene(xml_path, merge_shapes=False)
scene.frequency = CARRIER_FREQ
print(f"Scene loaded: {len(scene.objects)} objects")
print(f"Sample materials: {[o.radio_material.name for o in list(scene.objects.values())[:4]]}")
# Single isotropic antenna for both TX and RX
single_ant = PlanarArray(
num_rows=1,
num_cols=1,
vertical_spacing=0.5,
horizontal_spacing=0.5,
pattern="iso",
polarization="V",
)
scene.tx_array = single_ant
scene.rx_array = single_ant
Scene loaded: 266 objects Sample materials: ['concrete', 'concrete', 'concrete', 'concrete']
Step 3 — Place Transmitter and Receivers¶
The base station is placed at the center of the area at rooftop height. UEs are placed on a regular grid at street level.
In [8]:
Copied!
origin_lat, origin_lon = get_origin_coords(osm_scene_folder)
print(f"Local coordinate origin: ({origin_lat:.6f}, {origin_lon:.6f})")
xmin, ymin, xmax, ymax = convert_GpsBBox2CartesianBBox(
BBOX["minlat"],
BBOX["minlon"],
BBOX["maxlat"],
BBOX["maxlon"],
origin_lat,
origin_lon,
)
print(f"Scene extent: X=[{xmin:.0f}, {xmax:.0f}] m, Y=[{ymin:.0f}, {ymax:.0f}] m")
# Add transmitter
scene.add(Transmitter("tx_0", position=TX_POS[0].tolist()))
print(f"TX at {TX_POS[0]}")
# Generate UE grid
rx_pos = gen_plane_grid(xmin + 10, xmax - 10, ymin + 10, ymax - 10, GRID_SPACING, UE_HEIGHT)
print(f"UE grid: {len(rx_pos)} positions")
for i, pos in enumerate(rx_pos):
rx = Receiver(f"rx_{i}", position=pos.tolist())
rx.display_radius = 5.0 # metres; default heuristic is too large for this scene
scene.add(rx)
print(f"Added {len(rx_pos)} receivers")
origin_lat, origin_lon = get_origin_coords(osm_scene_folder)
print(f"Local coordinate origin: ({origin_lat:.6f}, {origin_lon:.6f})")
xmin, ymin, xmax, ymax = convert_GpsBBox2CartesianBBox(
BBOX["minlat"],
BBOX["minlon"],
BBOX["maxlat"],
BBOX["maxlon"],
origin_lat,
origin_lon,
)
print(f"Scene extent: X=[{xmin:.0f}, {xmax:.0f}] m, Y=[{ymin:.0f}, {ymax:.0f}] m")
# Add transmitter
scene.add(Transmitter("tx_0", position=TX_POS[0].tolist()))
print(f"TX at {TX_POS[0]}")
# Generate UE grid
rx_pos = gen_plane_grid(xmin + 10, xmax - 10, ymin + 10, ymax - 10, GRID_SPACING, UE_HEIGHT)
print(f"UE grid: {len(rx_pos)} positions")
for i, pos in enumerate(rx_pos):
rx = Receiver(f"rx_{i}", position=pos.tolist())
rx.display_radius = 5.0 # metres; default heuristic is too large for this scene
scene.add(rx)
print(f"Added {len(rx_pos)} receivers")
Local coordinate origin: (48.137500, 11.576500) Scene extent: X=[-216, 216] m, Y=[-230, 230] m TX at [ 0. 0. 25.] UE grid: 240 positions Added 240 receivers
Visualize — Sionna RT Scene Render¶
Sionna renders the 3D scene using Mitsuba's physically-based renderer. Buildings are visible as 3D blocks; the red marker is the TX and the blue spheres are the UE grid.
In [9]:
Copied!
# Camera offset slightly south so Sionna's Z-up convention renders
# East (x+) right and North (y+) up — matching the 2D footprint above.
cam_top = Camera(position=[0.0, -5.0, 500.0], look_at=[0.0, 0.0, 0.0])
fig = scene.render(camera=cam_top, show_devices=True)
fig.suptitle("OSM Scene in Sionna RT — Top View (x = East →, y = North ↑)")
plt.show()
# Camera offset slightly south so Sionna's Z-up convention renders
# East (x+) right and North (y+) up — matching the 2D footprint above.
cam_top = Camera(position=[0.0, -5.0, 500.0], look_at=[0.0, 0.0, 0.0])
fig = scene.render(camera=cam_top, show_devices=True)
fig.suptitle("OSM Scene in Sionna RT — Top View (x = East →, y = North ↑)")
plt.show()
In [10]:
Copied!
# South-facing oblique: consistent with top-down (East right, North into scene)
cam_oblique = Camera(position=[-300.0, -300.0, 300.0], look_at=[0.0, 0.0, 30.0])
fig = scene.render(camera=cam_oblique, show_devices=True)
fig.suptitle("OSM Scene in Sionna RT — Perspective View")
plt.show()
# South-facing oblique: consistent with top-down (East right, North into scene)
cam_oblique = Camera(position=[-300.0, -300.0, 300.0], look_at=[0.0, 0.0, 30.0])
fig = scene.render(camera=cam_oblique, show_devices=True)
fig.suptitle("OSM Scene in Sionna RT — Perspective View")
plt.show()
Step 4 — Run Sionna RT Ray Tracing¶
In [11]:
Copied!
RT_PARAMS = {
"max_depth": MAX_DEPTH,
"los": True,
"specular_reflection": True,
"diffuse_reflection": False,
"refraction": False,
"samples_per_src": N_SAMPLES,
}
print("Running ray tracing...")
p_solver = PathSolver()
paths = p_solver(scene=scene, **RT_PARAMS)
print(f"Path delays shape (tau): {paths.tau.shape}")
print(" (num_rx, num_tx, num_paths)")
RT_PARAMS = {
"max_depth": MAX_DEPTH,
"los": True,
"specular_reflection": True,
"diffuse_reflection": False,
"refraction": False,
"samples_per_src": N_SAMPLES,
}
print("Running ray tracing...")
p_solver = PathSolver()
paths = p_solver(scene=scene, **RT_PARAMS)
print(f"Path delays shape (tau): {paths.tau.shape}")
print(" (num_rx, num_tx, num_paths)")
Running ray tracing...
Path delays shape (tau): (240, 1, 18) (num_rx, num_tx, num_paths)
Step 5 — Export with sionna_exporter¶
In [12]:
Copied!
rt_save_folder = str(WORK_DIR / "sionna_rt_export")
sionna_exporter(scene, paths, RT_PARAMS, rt_save_folder)
print(f"\nExported files in {rt_save_folder}:")
for f in sorted(Path(rt_save_folder).iterdir()):
print(f" {f.name:<40} {f.stat().st_size / 1024:7.1f} KB")
rt_save_folder = str(WORK_DIR / "sionna_rt_export")
sionna_exporter(scene, paths, RT_PARAMS, rt_save_folder)
print(f"\nExported files in {rt_save_folder}:")
for f in sorted(Path(rt_save_folder).iterdir()):
print(f" {f.name:<40} {f.stat().st_size / 1024:7.1f} KB")
Exported files in /home/joao/.cache/deepmimo/osm_pipeline_munich/sionna_rt_export: sionna_faces.pkl 72.7 KB sionna_material_indices.pkl 2.2 KB sionna_materials.pkl 0.4 KB sionna_objects.pkl 5.8 KB sionna_paths.pkl 256.5 KB sionna_rt_params.pkl 0.6 KB sionna_vertices.pkl 52.7 KB
Step 6 — Convert to DeepMIMO Format¶
In [13]:
Copied!
scenario_name = dm.convert(rt_save_folder, overwrite=True)
print(f"Converted scenario: {scenario_name}")
scenario_name = dm.convert(rt_save_folder, overwrite=True)
print(f"Converted scenario: {scenario_name}")
Determining converter... Using Sionna RT converter converting from sionna RT
Processing receivers for TX 0, Ant 0: 0%| | 0/240 [00:00<?, ?it/s]
Processing receivers for TX 0, Ant 0: 100%|██████████| 240/240 [00:00<00:00, 23749.75it/s]
Converted scenario: sionna_rt_export
Step 7 — Load and Inspect the DeepMIMO Dataset¶
In [14]:
Copied!
dataset = dm.load(scenario_name)
print(dataset)
dataset = dm.load(scenario_name)
print(dataset)
Loading TXRX PAIR: TXset 0 (tx_idx 0) & RXset 1 (rx_idxs 240)
{'aoa_az': array([[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
...,
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan]],
shape=(240, 18), dtype=float32),
'aoa_el': array([[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
...,
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan]],
shape=(240, 18), dtype=float32),
'aod_az': array([[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
...,
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan]],
shape=(240, 18), dtype=float32),
'aod_el': array([[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
...,
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan]],
shape=(240, 18), dtype=float32),
'delay': array([[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
...,
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan]],
shape=(240, 18), dtype=float32),
'inter': array([[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
...,
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan]],
shape=(240, 18), dtype=float32),
'inter_pos': array([[[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
...,
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]]],
[[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
...,
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]]],
[[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
...,
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]]],
...,
[[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
...,
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]]],
[[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
...,
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]]],
[[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
...,
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]],
[[nan, nan, nan],
[nan, nan, nan]]]], shape=(240, 18, 2, 3), dtype=float32),
'load_params': {'compat_v3': False,
'matrices': 'all',
'max_paths': 25,
'rx_sets': {1: array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12,
13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25,
26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38,
39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,
52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64,
65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77,
78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90,
91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103,
104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116,
117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129,
130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142,
143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155,
156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168,
169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181,
182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194,
195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207,
208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220,
221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233,
234, 235, 236, 237, 238, 239])},
'tx_sets': {0: array([0])}},
'materials': MaterialList(1 materials, names=['material_0']),
'name': 'sionna_rt_export',
'phase': array([[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
...,
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan]],
shape=(240, 18), dtype=float32),
'power': array([[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
...,
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan],
[nan, nan, nan, ..., nan, nan, nan]],
shape=(240, 18), dtype=float32),
'rt_params': {'diffuse_diffractions': 0,
'diffuse_final_interaction_only': True,
'diffuse_random_phases': True,
'diffuse_reflections': 1,
'diffuse_transmissions': 0,
'frequency': 3500000000,
'gps_bbox': [0, 0, 0, 0],
'max_diffractions': 0,
'max_path_depth': 2,
'max_reflections': 2,
'max_scattering': 0,
'max_transmissions': 0,
'num_rays': 1000000,
'raw_params': {'bandwidth': [1000000.0],
'diffraction': False,
'diffuse_reflection': False,
'frequency': [3500000000.0],
'los': True,
'max_depth': 2,
'max_num_paths_per_src': 1000000,
'num_samples': 1000000,
'raytracer_version': '2.0.1',
'reflection': True,
'refraction': False,
'rx_array_ant_pos': [[0.0], [0.0], [0.0]],
'rx_array_num_ant': 1,
'rx_array_size': 1,
'samples_per_src': 1000000,
'scattering': False,
'seed': 42,
'specular_reflection': True,
'synthetic_array': True,
'tx_array_ant_pos': [[0.0], [0.0], [0.0]],
'tx_array_num_ant': 1,
'tx_array_size': 1},
'ray_casting_method': 'uniform',
'ray_casting_range_az': 360,
'ray_casting_range_el': 180,
'raytracer_name': 'Sionna Ray Tracing',
'raytracer_version': '2.0.1',
'synthetic_array': True,
'terrain_diffraction': False,
'terrain_reflection': True,
'terrain_scattering': False},
'rx_pos': array([[-205.72656 , -219.74727 , 1.5 ],
[-175.72656 , -219.74727 , 1.5 ],
[-145.72656 , -219.74727 , 1.5 ],
[-115.72656 , -219.74727 , 1.5 ],
[ -85.72656 , -219.74727 , 1.5 ],
[ -55.72656 , -219.74727 , 1.5 ],
[ -25.72656 , -219.74727 , 1.5 ],
[ 4.2734404, -219.74727 , 1.5 ],
[ 34.27344 , -219.74727 , 1.5 ],
[ 64.27344 , -219.74727 , 1.5 ],
[ 94.27344 , -219.74727 , 1.5 ],
[ 124.27344 , -219.74727 , 1.5 ],
[ 154.27344 , -219.74727 , 1.5 ],
[ 184.27344 , -219.74727 , 1.5 ],
[ 214.27344 , -219.74727 , 1.5 ],
[-205.72656 , -189.74727 , 1.5 ],
[-175.72656 , -189.74727 , 1.5 ],
[-145.72656 , -189.74727 , 1.5 ],
[-115.72656 , -189.74727 , 1.5 ],
[ -85.72656 , -189.74727 , 1.5 ],
[ -55.72656 , -189.74727 , 1.5 ],
[ -25.72656 , -189.74727 , 1.5 ],
[ 4.2734404, -189.74727 , 1.5 ],
[ 34.27344 , -189.74727 , 1.5 ],
[ 64.27344 , -189.74727 , 1.5 ],
[ 94.27344 , -189.74727 , 1.5 ],
[ 124.27344 , -189.74727 , 1.5 ],
[ 154.27344 , -189.74727 , 1.5 ],
[ 184.27344 , -189.74727 , 1.5 ],
[ 214.27344 , -189.74727 , 1.5 ],
[-205.72656 , -159.74727 , 1.5 ],
[-175.72656 , -159.74727 , 1.5 ],
[-145.72656 , -159.74727 , 1.5 ],
[-115.72656 , -159.74727 , 1.5 ],
[ -85.72656 , -159.74727 , 1.5 ],
[ -55.72656 , -159.74727 , 1.5 ],
[ -25.72656 , -159.74727 , 1.5 ],
[ 4.2734404, -159.74727 , 1.5 ],
[ 34.27344 , -159.74727 , 1.5 ],
[ 64.27344 , -159.74727 , 1.5 ],
[ 94.27344 , -159.74727 , 1.5 ],
[ 124.27344 , -159.74727 , 1.5 ],
[ 154.27344 , -159.74727 , 1.5 ],
[ 184.27344 , -159.74727 , 1.5 ],
[ 214.27344 , -159.74727 , 1.5 ],
[-205.72656 , -129.74727 , 1.5 ],
[-175.72656 , -129.74727 , 1.5 ],
[-145.72656 , -129.74727 , 1.5 ],
[-115.72656 , -129.74727 , 1.5 ],
[ -85.72656 , -129.74727 , 1.5 ],
[ -55.72656 , -129.74727 , 1.5 ],
[ -25.72656 , -129.74727 , 1.5 ],
[ 4.2734404, -129.74727 , 1.5 ],
[ 34.27344 , -129.74727 , 1.5 ],
[ 64.27344 , -129.74727 , 1.5 ],
[ 94.27344 , -129.74727 , 1.5 ],
[ 124.27344 , -129.74727 , 1.5 ],
[ 154.27344 , -129.74727 , 1.5 ],
[ 184.27344 , -129.74727 , 1.5 ],
[ 214.27344 , -129.74727 , 1.5 ],
[-205.72656 , -99.74727 , 1.5 ],
[-175.72656 , -99.74727 , 1.5 ],
[-145.72656 , -99.74727 , 1.5 ],
[-115.72656 , -99.74727 , 1.5 ],
[ -85.72656 , -99.74727 , 1.5 ],
[ -55.72656 , -99.74727 , 1.5 ],
[ -25.72656 , -99.74727 , 1.5 ],
[ 4.2734404, -99.74727 , 1.5 ],
[ 34.27344 , -99.74727 , 1.5 ],
[ 64.27344 , -99.74727 , 1.5 ],
[ 94.27344 , -99.74727 , 1.5 ],
[ 124.27344 , -99.74727 , 1.5 ],
[ 154.27344 , -99.74727 , 1.5 ],
[ 184.27344 , -99.74727 , 1.5 ],
[ 214.27344 , -99.74727 , 1.5 ],
[-205.72656 , -69.74727 , 1.5 ],
[-175.72656 , -69.74727 , 1.5 ],
[-145.72656 , -69.74727 , 1.5 ],
[-115.72656 , -69.74727 , 1.5 ],
[ -85.72656 , -69.74727 , 1.5 ],
[ -55.72656 , -69.74727 , 1.5 ],
[ -25.72656 , -69.74727 , 1.5 ],
[ 4.2734404, -69.74727 , 1.5 ],
[ 34.27344 , -69.74727 , 1.5 ],
[ 64.27344 , -69.74727 , 1.5 ],
[ 94.27344 , -69.74727 , 1.5 ],
[ 124.27344 , -69.74727 , 1.5 ],
[ 154.27344 , -69.74727 , 1.5 ],
[ 184.27344 , -69.74727 , 1.5 ],
[ 214.27344 , -69.74727 , 1.5 ],
[-205.72656 , -39.74727 , 1.5 ],
[-175.72656 , -39.74727 , 1.5 ],
[-145.72656 , -39.74727 , 1.5 ],
[-115.72656 , -39.74727 , 1.5 ],
[ -85.72656 , -39.74727 , 1.5 ],
[ -55.72656 , -39.74727 , 1.5 ],
[ -25.72656 , -39.74727 , 1.5 ],
[ 4.2734404, -39.74727 , 1.5 ],
[ 34.27344 , -39.74727 , 1.5 ],
[ 64.27344 , -39.74727 , 1.5 ],
[ 94.27344 , -39.74727 , 1.5 ],
[ 124.27344 , -39.74727 , 1.5 ],
[ 154.27344 , -39.74727 , 1.5 ],
[ 184.27344 , -39.74727 , 1.5 ],
[ 214.27344 , -39.74727 , 1.5 ],
[-205.72656 , -9.74727 , 1.5 ],
[-175.72656 , -9.74727 , 1.5 ],
[-145.72656 , -9.74727 , 1.5 ],
[-115.72656 , -9.74727 , 1.5 ],
[ -85.72656 , -9.74727 , 1.5 ],
[ -55.72656 , -9.74727 , 1.5 ],
[ -25.72656 , -9.74727 , 1.5 ],
[ 4.2734404, -9.74727 , 1.5 ],
[ 34.27344 , -9.74727 , 1.5 ],
[ 64.27344 , -9.74727 , 1.5 ],
[ 94.27344 , -9.74727 , 1.5 ],
[ 124.27344 , -9.74727 , 1.5 ],
[ 154.27344 , -9.74727 , 1.5 ],
[ 184.27344 , -9.74727 , 1.5 ],
[ 214.27344 , -9.74727 , 1.5 ],
[-205.72656 , 20.252731 , 1.5 ],
[-175.72656 , 20.252731 , 1.5 ],
[-145.72656 , 20.252731 , 1.5 ],
[-115.72656 , 20.252731 , 1.5 ],
[ -85.72656 , 20.252731 , 1.5 ],
[ -55.72656 , 20.252731 , 1.5 ],
[ -25.72656 , 20.252731 , 1.5 ],
[ 4.2734404, 20.252731 , 1.5 ],
[ 34.27344 , 20.252731 , 1.5 ],
[ 64.27344 , 20.252731 , 1.5 ],
[ 94.27344 , 20.252731 , 1.5 ],
[ 124.27344 , 20.252731 , 1.5 ],
[ 154.27344 , 20.252731 , 1.5 ],
[ 184.27344 , 20.252731 , 1.5 ],
[ 214.27344 , 20.252731 , 1.5 ],
[-205.72656 , 50.25273 , 1.5 ],
[-175.72656 , 50.25273 , 1.5 ],
[-145.72656 , 50.25273 , 1.5 ],
[-115.72656 , 50.25273 , 1.5 ],
[ -85.72656 , 50.25273 , 1.5 ],
[ -55.72656 , 50.25273 , 1.5 ],
[ -25.72656 , 50.25273 , 1.5 ],
[ 4.2734404, 50.25273 , 1.5 ],
[ 34.27344 , 50.25273 , 1.5 ],
[ 64.27344 , 50.25273 , 1.5 ],
[ 94.27344 , 50.25273 , 1.5 ],
[ 124.27344 , 50.25273 , 1.5 ],
[ 154.27344 , 50.25273 , 1.5 ],
[ 184.27344 , 50.25273 , 1.5 ],
[ 214.27344 , 50.25273 , 1.5 ],
[-205.72656 , 80.25273 , 1.5 ],
[-175.72656 , 80.25273 , 1.5 ],
[-145.72656 , 80.25273 , 1.5 ],
[-115.72656 , 80.25273 , 1.5 ],
[ -85.72656 , 80.25273 , 1.5 ],
[ -55.72656 , 80.25273 , 1.5 ],
[ -25.72656 , 80.25273 , 1.5 ],
[ 4.2734404, 80.25273 , 1.5 ],
[ 34.27344 , 80.25273 , 1.5 ],
[ 64.27344 , 80.25273 , 1.5 ],
[ 94.27344 , 80.25273 , 1.5 ],
[ 124.27344 , 80.25273 , 1.5 ],
[ 154.27344 , 80.25273 , 1.5 ],
[ 184.27344 , 80.25273 , 1.5 ],
[ 214.27344 , 80.25273 , 1.5 ],
[-205.72656 , 110.25273 , 1.5 ],
[-175.72656 , 110.25273 , 1.5 ],
[-145.72656 , 110.25273 , 1.5 ],
[-115.72656 , 110.25273 , 1.5 ],
[ -85.72656 , 110.25273 , 1.5 ],
[ -55.72656 , 110.25273 , 1.5 ],
[ -25.72656 , 110.25273 , 1.5 ],
[ 4.2734404, 110.25273 , 1.5 ],
[ 34.27344 , 110.25273 , 1.5 ],
[ 64.27344 , 110.25273 , 1.5 ],
[ 94.27344 , 110.25273 , 1.5 ],
[ 124.27344 , 110.25273 , 1.5 ],
[ 154.27344 , 110.25273 , 1.5 ],
[ 184.27344 , 110.25273 , 1.5 ],
[ 214.27344 , 110.25273 , 1.5 ],
[-205.72656 , 140.25273 , 1.5 ],
[-175.72656 , 140.25273 , 1.5 ],
[-145.72656 , 140.25273 , 1.5 ],
[-115.72656 , 140.25273 , 1.5 ],
[ -85.72656 , 140.25273 , 1.5 ],
[ -55.72656 , 140.25273 , 1.5 ],
[ -25.72656 , 140.25273 , 1.5 ],
[ 4.2734404, 140.25273 , 1.5 ],
[ 34.27344 , 140.25273 , 1.5 ],
[ 64.27344 , 140.25273 , 1.5 ],
[ 94.27344 , 140.25273 , 1.5 ],
[ 124.27344 , 140.25273 , 1.5 ],
[ 154.27344 , 140.25273 , 1.5 ],
[ 184.27344 , 140.25273 , 1.5 ],
[ 214.27344 , 140.25273 , 1.5 ],
[-205.72656 , 170.25273 , 1.5 ],
[-175.72656 , 170.25273 , 1.5 ],
[-145.72656 , 170.25273 , 1.5 ],
[-115.72656 , 170.25273 , 1.5 ],
[ -85.72656 , 170.25273 , 1.5 ],
[ -55.72656 , 170.25273 , 1.5 ],
[ -25.72656 , 170.25273 , 1.5 ],
[ 4.2734404, 170.25273 , 1.5 ],
[ 34.27344 , 170.25273 , 1.5 ],
[ 64.27344 , 170.25273 , 1.5 ],
[ 94.27344 , 170.25273 , 1.5 ],
[ 124.27344 , 170.25273 , 1.5 ],
[ 154.27344 , 170.25273 , 1.5 ],
[ 184.27344 , 170.25273 , 1.5 ],
[ 214.27344 , 170.25273 , 1.5 ],
[-205.72656 , 200.25273 , 1.5 ],
[-175.72656 , 200.25273 , 1.5 ],
[-145.72656 , 200.25273 , 1.5 ],
[-115.72656 , 200.25273 , 1.5 ],
[ -85.72656 , 200.25273 , 1.5 ],
[ -55.72656 , 200.25273 , 1.5 ],
[ -25.72656 , 200.25273 , 1.5 ],
[ 4.2734404, 200.25273 , 1.5 ],
[ 34.27344 , 200.25273 , 1.5 ],
[ 64.27344 , 200.25273 , 1.5 ],
[ 94.27344 , 200.25273 , 1.5 ],
[ 124.27344 , 200.25273 , 1.5 ],
[ 154.27344 , 200.25273 , 1.5 ],
[ 184.27344 , 200.25273 , 1.5 ],
[ 214.27344 , 200.25273 , 1.5 ],
[-205.72656 , 230.25273 , 1.5 ],
[-175.72656 , 230.25273 , 1.5 ],
[-145.72656 , 230.25273 , 1.5 ],
[-115.72656 , 230.25273 , 1.5 ],
[ -85.72656 , 230.25273 , 1.5 ],
[ -55.72656 , 230.25273 , 1.5 ],
[ -25.72656 , 230.25273 , 1.5 ],
[ 4.2734404, 230.25273 , 1.5 ],
[ 34.27344 , 230.25273 , 1.5 ],
[ 64.27344 , 230.25273 , 1.5 ],
[ 94.27344 , 230.25273 , 1.5 ],
[ 124.27344 , 230.25273 , 1.5 ],
[ 154.27344 , 230.25273 , 1.5 ],
[ 184.27344 , 230.25273 , 1.5 ],
[ 214.27344 , 230.25273 , 1.5 ]], dtype=float32),
'scene': Scene(266 objects [terrain: 1, buildings: 265], dims = 555.2 x 537.3 x 91.0 m),
'tx_pos': array([[ 0., 0., 25.]], dtype=float32),
'txrx': {'rx_set_id': 1, 'tx_idx': 0, 'tx_set_id': 0}}
Step 8 — Received Power Coverage Map¶
Per-UE peak received power (across paths) plotted as a scatter map. UEs with no paths (obstructed or out-of-range) are shown in grey.
In [15]:
Copied!
rx_pos_arr = np.array(dataset.rx_pos)
power_peak = np.nanmax(dataset.power, axis=1) # (n_ue,) in dBW
fig, ax = plt.subplots(figsize=(8, 7))
sc = ax.scatter(
rx_pos_arr[:, 0],
rx_pos_arr[:, 1],
c=power_peak,
cmap="viridis",
s=60,
vmin=np.nanpercentile(power_peak, 5),
)
ax.scatter(*TX_POS[0, :2], c="red", marker="^", s=200, zorder=5, label="TX")
plt.colorbar(sc, ax=ax, label="Peak received power (dBW)")
ax.set_xlabel("x (m)")
ax.set_ylabel("y (m)")
ax.set_title(f"Coverage Map — OSM Munich ({len(rx_pos)} UEs, {CARRIER_FREQ / 1e9:.1f} GHz)")
ax.legend()
ax.grid(visible=True, alpha=0.3)
plt.tight_layout()
plt.show()
rx_pos_arr = np.array(dataset.rx_pos)
power_peak = np.nanmax(dataset.power, axis=1) # (n_ue,) in dBW
fig, ax = plt.subplots(figsize=(8, 7))
sc = ax.scatter(
rx_pos_arr[:, 0],
rx_pos_arr[:, 1],
c=power_peak,
cmap="viridis",
s=60,
vmin=np.nanpercentile(power_peak, 5),
)
ax.scatter(*TX_POS[0, :2], c="red", marker="^", s=200, zorder=5, label="TX")
plt.colorbar(sc, ax=ax, label="Peak received power (dBW)")
ax.set_xlabel("x (m)")
ax.set_ylabel("y (m)")
ax.set_title(f"Coverage Map — OSM Munich ({len(rx_pos)} UEs, {CARRIER_FREQ / 1e9:.1f} GHz)")
ax.legend()
ax.grid(visible=True, alpha=0.3)
plt.tight_layout()
plt.show()
/tmp/ipykernel_656302/2146066148.py:2: RuntimeWarning: All-NaN axis encountered power_peak = np.nanmax(dataset.power, axis=1) # (n_ue,) in dBW
Summary¶
| Step | Tool | Output |
|---|---|---|
| 1. OSM → scene | generate_scene |
scene.xml + PLY meshes |
| 2. Load scene | sionna.rt.load_scene |
Sionna Scene object |
| 3. Place antennas | Transmitter / Receiver |
TX + UE grid |
| 4. Ray trace | PathSolver |
Paths object |
| 5. Export | sionna_exporter |
.pkl files |
| 6. Convert | dm.convert |
DeepMIMO scenario |
| 7. Load | dm.load |
Dataset object |
| 8. Coverage map | dataset.power |
Per-UE received power scatter |
Key design choices:
- No Blender —
generate_scenefetches OSM data via the Overpass API and writes Mitsuba PLY meshes directly in Python. - ITU radio materials — buildings use
itu-radio-material type=concrete, matching the material convention of Sionna's built-in scenes. - Local coordinate system — the center of the GPS bounding box is the
origin;
osm_gps_origin.txtrecords this for downstream use. - Reusable scene folder — the generated
scene.xmlfolder can be fed directly intoraytrace_sionna()for large-scale batch processing. - Satellite overlay — optionally add a Google Maps satellite image with
deepmimo.pipelines.utils.geo_utils.fetch_satellite_view(bbox, api_key, ...).