RTT-LIO: A Wi-Fi RTT-aided LiDAR-Inertial Odometry via Tightly-Coupled Factor Graph Optimization for UAS

RTT-LIO is a novel tightly-coupled factor graph optimization framework that integrates Wi-Fi Round-Trip Time (RTT) measurements with LiDAR and inertial data for robust state estimation in Unmanned Aerial Systems (UAS). This system addresses the critical challenges of accurate localization in GPS-denied environments by using the complementary characteristics of multiple sensing modalities. The package is based on C++ which is compatible with the robot operation system (ROS) platform.

Authors: Ruijie Xu Xikun Liu*, Xin Wang, Weisong Wen, from the TASLAB, The Hong Kong Polytechnic University. And Yulong Huang, from HEU

Contact: For questions, bug reports, or collaboration inquiries, please contact ruijie.xu@connect.polyu.hk.

Citation: If you use this repository, please cite our paper:

@article{xu2026rtt,
  title={RTT-LIO: A Wi-Fi RTT-aided LiDAR-Inertial Odometry via Tightly-Coupled Factor Graph Optimization in Complex Scenes},
  author={Xu, Ruijie and Liu, Xikun and Wang, Xin and Wen, Weisong and Huang, Yulong},
  journal={IEEE Internet of Things Journal},
  year={2026},
  publisher={IEEE}
}

System Pipeline

RTT-LIO implements a comprehensive multi-stage processing pipeline:

1. Frontend Processing:
   • Keyframe Selection and Initialization: Selects informative LiDAR frames for processing and establishes system initialization parameters
   • IMU Pre-integration: Performs integration of inertial measurements between LiDAR frames to capture high-frequency motion dynamics
   • RTT Range from Multi-APs Modeling: Models the ranging measurements from multiple Wi-Fi access points, handling outliers and non-line-of-sight conditions
2. Factor Graph Construction:
   • Scan-to-multiscan LiDAR Factor: Establishes geometric constraints between consecutive LiDAR scans
   • IMU Factor: Provides motion constraints from inertial measurements
3. Two-Stage Optimization:
   • Global Pose from First-stage FGO: Performs initial optimization to estimate UAS trajectory
   • RTT Range Factor: Adds distance constraints between the UAS and Wi-Fi access points
   • Batch-based Second-stage FGO with FDE: Implements Fault Detection and Exclusion (FDE) to identify and reject erroneous measurements, enhancing trajectory estimation robustness
4. AP Position Calibration (Optional):
   • RANSAC-based KDE and GDC: Uses Random Sample Consensus with Kernel Density Estimation and Geometric Diversity Constraints for accurate access point position estimation, leveraging the estimated UAS trajectory

This tightly-coupled architecture enables RTT-LIO to maintain submeter-level accuracy even in challenging environments by dynamically weighting measurements based on their reliability and ensuring consistent constraints across the entire trajectory.

Code and Datasets

Three ROS nodes run in sequence:

Node	Role
Preprocessing	Subscribes to raw LiDAR scans, extracts edge/surface features
LidarOdometry	Scan-to-scan feature matching to produce a high-rate odometry prior
Estimator	Sliding-window + batch factor graph (IMU pre-integration, LiDAR pose, Wi-Fi RTT range)

---

Key Features

• Tightly-coupled Wi-Fi RTT / LiDAR / IMU fusion in a single factor graph
• Sliding-window optimization (real-time) + batch re-optimization for global consistency
• Marginalization for bounded computational cost
• Loop closure via ICP + GTSAM iSAM2 (optional, toggle loop_closure_on in config)
• AP position agnostic: works with pre-surveyed AP coordinates
• Supports 16/32/64-beam Velodyne LiDARs

---

Prerequisites

Dependency	Version tested	Notes
Ubuntu	20.04
ROS	Noetic	full desktop
C++	14
CMake	≥ 3.10
Eigen	3.3.7 (system)	GTSAM bundles its own 3.4; handled automatically
PCL	1.10	via ros-noetic-pcl-ros
Ceres Solver	2.2.0	build from source
GTSAM	4.3	build from source (-DGTSAM_USE_SYSTEM_EIGEN=OFF)

---

Installation

Option A — Docker (recommended)

The provided Dockerfile installs all dependencies and builds the workspace automatically.

# Clone the repo
git clone https://github.com/RuijieXu0408/RTT-LIO.git -b rttlio-release
cd RTT-LIO

# Build the image (takes ~10–15 min on first run)
docker build -f src/RTTLIO/Dockerfile -t rttlio:latest .

# Run with display forwarding (for RViz)
xhost +local:docker
docker run -it --rm \
    --net=host \
    -e DISPLAY=$DISPLAY \
    -v /tmp/.X11-unix:/tmp/.X11-unix \
    -v /path/to/your/bag:/data \
    rttlio:latest bash

Inside the container, the workspace is at /root/rttlio_ws and the environment is already sourced.

---

Option B — Native build

1. Install ROS Noetic

Follow the official guide.

sudo apt install ros-noetic-desktop-full

2. Install system dependencies

sudo apt install -y \
    cmake build-essential git wget \
    libatlas-base-dev libgoogle-glog-dev libsuitesparse-dev \
    libpcl-dev \
    ros-noetic-pcl-ros ros-noetic-pcl-conversions \
    ros-noetic-velodyne-msgs

3. Build Ceres Solver 2.2.0

git clone --depth 1 --branch 2.2.0 \
    https://ceres-solver.googlesource.com/ceres-solver
mkdir ceres-bin && cd ceres-bin
cmake ../ceres-solver \
    -DCMAKE_BUILD_TYPE=Release \
    -DBUILD_TESTING=OFF \
    -DBUILD_EXAMPLES=OFF
make -j$(nproc) && sudo make install

4. Build GTSAM 4.3

Build with the bundled Eigen (-DGTSAM_USE_SYSTEM_EIGEN=OFF) to avoid the Eigen version assertion failure.

git clone --depth 1 --branch 4.3a0 https://github.com/borglab/gtsam.git
mkdir gtsam-build && cd gtsam-build
cmake ../gtsam \
    -DCMAKE_BUILD_TYPE=Release \
    -DGTSAM_BUILD_TESTS=OFF \
    -DGTSAM_BUILD_EXAMPLES_ALWAYS=OFF \
    -DGTSAM_USE_SYSTEM_EIGEN=OFF
make -j$(nproc) && sudo make install

5. Create workspace and build RTT-LIO

mkdir -p ~/rttlio_ws/src && cd ~/rttlio_ws/src
git clone https://github.com/RuijieXu0408/RTT-LIO.git \
    --branch rttlio-release --single-branch rttlio_pkg

cd ~/rttlio_ws
source /opt/ros/noetic/setup.bash
catkin_make -DCMAKE_BUILD_TYPE=Release

---

Configuration

Edit src/RTTLIO/config/config_urban_hk.yaml before launching:

IMU:
  imu_topic: "/imu/data"        # change to match your bag
  acc_n: 3.99e-03               # accelerometer noise density
  gyr_n: 1.56e-03               # gyroscope noise density

lidar_odometry:
  lidar_topic: "/velodyne_points"
  line_num: 32                  # 16 / 32 / 64

initialization:
  Euler_y: -155                 # initial yaw (degrees) — tune per dataset
  gt_path: /path/to/gt.csv      # ground-truth CSV for evaluation (optional)

Estimator:
  enable_batch_fusion: true     # enable batch RTT re-optimization
  loop_closure_on: false        # enable loop closure (requires more CPU)
  slide_window_width: 5
  surf_dist_thres: 0.18

The Wi-Fi RTT data file is set in the launch file:

<!-- run_urban_hk.launch -->
<param name="sol_folder" type="string" value="/path/to/your/rtt_data.csv" />

---

Running

# Terminal 1 — source and launch
source ~/rttlio_ws/devel/setup.bash
roslaunch RTTLIO run_urban_hk.launch

# Terminal 2 — play your bag
rosbag play /path/to/your.bag

Expected topics the bag should contain:

Topic	Type	Description
/velodyne_points	sensor_msgs/PointCloud2	3D LiDAR scan
/imu/data	sensor_msgs/Imu	IMU (200 Hz recommended)

Wi-Fi RTT measurements are loaded from the CSV file specified by sol_folder (not from a ROS topic).

---

Dataset Format

The RTT data CSV (sol_folder) has one measurement per line:

unixtime, ap_index, ap_pos_x, ap_pos_y, ap_pos_z, range

Field	Unit	Description
unixtime	s	Unix timestamp of the RTT measurement
ap_index	—	Access point identifier (integer)
ap_pos_x/y/z	m	AP position in the local ENU frame
range	m	RTT-derived range measurement

Example datasets can be found in the WiFiRTT_Dataset/ folder (see the repository).