An open-source software platform for all-scenario wireless communication and sensing research.

---

SimART is an open-source software platform for all-scenario wireless communication and sensing research. Built around ROS1, it integrates a C++/Qt/VTK graphical interface, Sionna-based ray tracing and link simulation, AirSim/Unreal Engine live visualization, and rosbag recording and replay tools. It helps users place base stations, visualize trajectories, inspect wireless channel observations, and evaluate beam selection workflows in 3D scenes.

It supports digital-twin scene construction, ROS-based trajectory replay, Sionna data collection, visual network planning, and communication-sensing experiments with simulators such as AirSim and Unreal Engine.

Contents

• System Architecture
• Scene Construction and Map Adaptation
• Multimodal Data and CKM
• Key Features
• Preview
• Tutorials
• Quick Start
• Try SimART
• Use SimART with UE4 and AirSim
• Further Exploration

System Architecture

SimART consists of four functional modules coordinated by ROS:

Module	Role
Physics and Sensing Module	Provides platform motion, RGB/depth/semantic cameras, LiDAR, IMU, GPS, and ground-truth poses through ROS-compatible simulators such as AirSim, Gazebo, Isaac Sim, or CARLA.
Ray Tracing Module	Uses Sionna RT to compute site-specific propagation paths, delays, angles, Doppler shifts, interaction points, and channel impulse responses.
Link and System Module	Uses Sionna SYS to evaluate OFDM, PHY/MAC behavior, multi-antenna links, beamforming codebooks, SINR, BLER, achievable rate, and optimal beam index.
CKM Generator	Scans dense receiver grids to generate multi-layer channel knowledge maps for path loss, delay/angular spread, SINR, rate, and beam-selection priors.

The ray tracing module supports online simulation, offline rosbag replay, and dense grid scan modes. This allows users to visualize propagation during a live session, reproduce experiments from recorded trajectories, or generate CKM layers over a region of interest.

Scene Construction and Map Adaptation

SimART provides two complementary scene construction pipelines. Both produce a high-fidelity visual scene for physics/sensing and a simplified, material-aware scene for ray tracing, while preserving a shared coordinate frame.

Real-world map adaptation: OpenStreetMap data are converted into aligned visual and ray-tracing assets.

User-defined scene interface: RoadRunner or Unreal Engine scenes are simplified into propagation-ready meshes.

Pipeline	Workflow
Real-world map adaptation	OpenStreetMap extracts provide building footprints, heights, roads, and land-use polygons. OSM2World creates the visual asset for Unreal Engine/AirSim, while Blender exports a Mitsuba/Sionna RT scene with electromagnetic material annotations.
User-defined scene interface	RoadRunner, Unreal Engine, or other custom scene assets are imported for physics and sensing. A Blender conversion script removes fine visual details, applies mesh decimation, preserves dominant facades and ground planes, and assigns Sionna RT materials.

Multimodal Data and CKM

During a simulation session, every module publishes data under the shared ROS clock. A single rosbag can preserve the complete synchronized session for replay, inspection, dataset export, and downstream learning tasks.

Multimodal sensor streams: RGB, depth, LiDAR, and semantic segmentation.	Dense CKM layers: path loss, received power, best BS rate, and effective SINR.

Data Stream	Examples
Sensing and mobility	RGB images, depth images, semantic images, LiDAR point clouds, IMU, GPS, ground-truth poses, and navigation messages.
Wireless channel	Per-link CIRs, propagation paths, path loss, delays, angles of arrival/departure, Doppler, and interaction points.
Link/system KPIs	SINR, BLER, achievable rate, PHY/MAC outputs, and optimal beam index from a configured codebook.
CKM layers	Path loss, RMS delay spread, angular spread, average SINR, achievable rate, and beam-selection priors over dense spatial grids.

Key Features

Capability	Description
3D scene loading and preview	Load local scene meshes in the GUI to inspect maps, UAV trajectories, base stations, and ray paths.
Sionna wireless simulation	Run ray tracing, OFDM/SYS link adaptation, and beam codebook selection from exported Mitsuba/XML scenes.
ROS1 data-stream integration	Subscribe to UAV pose topics and publish RF, SYS, beam, codebook, and related simulation observations.
AirSim/UE live integration	Display UAVs, trajectories, base stations, and rays directly in Unreal Engine scenes, with support for base-station camera previews.
Base-station editing and configuration	Add, select, edit, save, and load base-station JSON configurations from the GUI.
Rosbag tools	Record, replay, and re-simulate wireless data for offline analysis and dataset generation.
Coordinate-frame configuration	Configure transforms among the ROS frame, 3D scene frame, and AirSim frame.

Preview

Main Simulation Interface	UAV Perspective	Base-Station Perspective

Tutorials

The following brings you a quick start. For other tutorials, click here:

Tutorial	Link
Create Your Own Maps	Tutorials/CreateMap/CreateMap.md
Use SimART with Your Own Maps	Tutorials/Usage/Usage.md

Quick Start

Building and Compiling

This project is a ROS 1 catkin workspace package set and was tested under Ubuntu 20.04 + ROS 1 Noetic. We recommend working with the same environment. The repository should be placed under a catkin workspace, for example:

mkdir -p ~/catkin_ws/src
cd ~/catkin_ws/src
git clone https://github.com/guchuanv-alt/SimART.git

The repository contains these catkin packages:

Package	Description
airsim_gui_UErealtime	the C++/Qt/VTK GUI package
rf_msgs	RF observation message definitions
sionna_sys_msgs	Sionna SYS message definitions
sionna_beam_msgs	Sionna beam/codebook message definitions

1. System and ROS Dependencies

Install ROS 1 Noetic for Ubuntu 20.04 first.

We recommend following the official ROS Noetic installation instructions for Ubuntu 20.04: http://wiki.ros.org/noetic/Installation/Ubuntu

If you prefer to use the third-party FishROS convenience installer, treat it as an optional alternative and use the HTTPS endpoint: https://github.com/fishros/install

wget https://fishros.com/install -O fishros && bash fishros

After ROS 1 Noetic is installed, source the ROS environment and install the remaining system dependencies:

source /opt/ros/noetic/setup.bash
sudo apt update
sudo apt install -y \
  build-essential \
  cmake \
  python3-pip \
  python3-venv \
  python3-rosdep \
  python3-catkin-tools \
  qtbase5-dev \
  qtdeclarative5-dev \
  libqt5opengl5-dev \
  qml-module-qtquick2 \
  libassimp-dev \
  libvtk7-dev

If your Ubuntu version does not provide libvtk7-dev, install the available VTK development package instead, for example:

sudo apt install -y libvtk9-dev

Then install ROS package dependencies declared by the catkin packages:

cd ~/catkin_ws
sudo rosdep init 2>/dev/null || true
rosdep update
rosdep install --from-paths src/SimART --ignore-src -r -y

rosdep can now install the Qt, VTK, and Assimp dependencies declared by SimART_GUI/package.xml. The explicit apt install command above is still listed so a new machine has the core tools before running rosdep.

2. Python Dependencies

We recommend installing the Python dependencies with conda. The tested conda environment is named SimART and uses Python 3.12:

cd ~/catkin_ws/src/SimART
conda create -n SimART python=3.12
conda activate SimART
python -m pip install --upgrade pip
python -m pip install -r requirements.txt

requirements.txt includes sionna. You can also refer to Sionna to install it.

3. Build the Default GUI

Build the default GUI with catkin build:

cd ~/catkin_ws
source /opt/ros/noetic/setup.bash
catkin init
catkin config --extend /opt/ros/noetic --cmake-args -DCMAKE_BUILD_TYPE=RelWithDebInfo
catkin build airsim_gui_UErealtime
source devel/setup.bash

Run SimART:

rosrun airsim_gui_UErealtime airsim_gui_UErealtime

Try SimART

4 sample maps and 1 sample rosbag for one of them are provided. The quick-start demo uses a UAV trajectory in BigCitySample. The sample rosbag already contains the UAV pose data recorded from AirSim, so it can be used directly for simulation without installing AirSim first. The rest of the maps can be used for further exploration.

4. Download the Sample Maps and Rosbags

In the root directory of the repository you just cloned, run:

cd ~/catkin_ws/src/SimART
chmod +x download_sample_maps.sh
./download_sample_maps.sh

and:

cd ~/catkin_ws/src/SimART
chmod +x download_sample_rosbags.sh
./download_sample_rosbags.sh

The folder SimART_sample_maps contains 4 sample maps.
The folder SimART_sample_rosbags contains 1 rosbag for the map BigCitySample.

5. Load the Config File in SimART and Start the Simulation

Run SimART by using the command:

rosrun airsim_gui_UErealtime airsim_gui_UErealtime

• In SimART, click "Open Existing Config" and select ~/catkin_ws/src/SimART/SimART_sample_maps/BigCitySample/BigCitySample.agcfg.
• Click "Simulation Settings", select the SimART conda environment in the Python environment field, and click "Test Environment".

• Click "Open Rosbag Tools", select the downloaded sample rosbag in Replay panel. Click "Start Playback".

• Click "Start Simulation" to start the Sionna simulation. RF, SYS, and beam simulation will all be started. The data can be viewed in the Wireless Data and Sionna SYS panels. The raw data is available on ROS topics and can be inspected with the rostopic CLI.

• You can also record the generated data after clicking "Start Recording".

Use SimART with UE4 and AirSim

If you need to view the scene in UE4 or customize UAV flight, follow the steps below to set up Unreal Engine and AirSim.

6. Download and Compile Unreal Engine

Clone and build Unreal Engine 4 and AirSim. We recommend you to use Unreal Engine 4.27 version. You can refer to Unreal Engine Documentation and AirSim.

Make sure that you can use AirSim to perform UAV simulations first. Then move the AirSim ros wrapper package to your AirSim path. In the root directory of the repository, run (Remember to replace the path/to/your/AirSim with your actual AirSim path):

cp -r ~/catkin_ws/src/SimART/third_party/airsim_ros_pkgs_sa path/to/your/AirSim/ros/src

This will copy the modified AirSim ros wrapper package to the AirSim ros workspace. Then build the new AirSim ros wrapper in the root directory of the AirSim ros workspace:

cd path/to/your/AirSim/ros
catkin build

When you use the new ros wrapper to publish the rostopics in AirSim, you can run:

cd path/to/your/AirSim/ros
source devel/setup.bash
roslaunch airsim_ros_pkgs_sa airsim_node.launch is_vulkan:="false"

7. Enable UE Live View

The normal build can load local meshes and use ROS topics without the AirSim C++ SDK. Enable AirSim C++ support only when you need UE live-view integration(The UE Live View panel).

First make sure that Unreal Engine 4 and AirSim have been cloned and built, including AirSim's RPC library. Then rebuild the catkin workspace with(Remember to replace /path/to/your/AirSim with your actual AirSim path)

cd ~/catkin_ws
source /opt/ros/noetic/setup.bash
catkin build airsim_gui_UErealtime --cmake-args \
  -DAIRSIM_GUI_ENABLE_AIRSIM=ON \
  -DAIRSIM_CLIENT_ROOT=/path/to/your/AirSim
source devel/setup.bash

8. Further Exploration

The rest of the maps should work with a UAV simulation software, e.g., AirSim (Recommended), Gazebo, etc, and a matched map (Take AirSim for instance, a corresponding Unreal Engine project is required). The required output of the UAV simulation software is a rostopic containing the pose of the UAV (data type is nav_msgs/Odometry or geometry_msgs/PoseStamped). If you decide to use AirSim, please follow Create your own maps.

---

Contact

For questions, feedback, or collaboration, please contact us at kangyan@std.uestc.edu.cn, yuqicao@std.uestc.edu.cn.

---

Citation

If you use SimART in your research, please cite our paper:

@article{yan2026simart,
  title={SimART: A Unified and Open Real-world Multimodal Simulation Platform for 6G Integrated Sensing and Communication},
  author={Yan, Kang and Cao, Yuqi and Li, Jiaqi and Xiang, Luping and Yang, Kun},
  journal={arXiv preprint arXiv:2605.13309},
  year={2026}
}