Skip to content

Latest commit

 

History

History
75 lines (58 loc) · 3.28 KB

README.md

File metadata and controls

75 lines (58 loc) · 3.28 KB

Experimental Comparison of Tightly and Loosely Coupled Kalman Filters for 6D Pose Estimation Using UWB and IMU Sensors(2024 ICCAS)

we present a novel approach for robot pose estimation in SE(3) using a tightly coupled integration of Ultra-Wideband (UWB) and Inertial Measurement Unit (IMU) data. Our method leverages Extended Kalman Filter (EKF) and Error-State Kalman Filter (ESKF) to accurately estimate the pose, including position and orientation, of a mobile robot in three-dimensional space. The tightly coupled approach ensures robust performance even in environments where UWB data quality is poor or in NLOS(Non-Line-of-Sight) conditions by utilizing raw UWB data and effectively fusing with IMU. The proposed method is validated through extensive simulations and real-world experiments, demonstrating significant improvements in accuracy and robustness compared to the loosely coupled method. Our results show that the tightly coupled approach, with its enhanced data fusion capability, provides superior z-axis estimation performance. This makes it particularly advantageous in applications requiring accurate SE(3) pose estimation.

6D pose estimation using tightly coupled UWB/IMU fusion using filtering method(EKF/ESKF/UKF/LIEKF)

  • Extended Kalman filter
  • Error-state Kalman filter
  • Unscented Kalman filter
  • Left-invariant Kalman filter

Experimental setup

  • SBC: Intel NUC-i7
  • UWB(50Hz): Nooploop LinkTrack P-b (1tag, 8anchors)
  • IMU(20Hz): UM7
  • Ground Truth data: Qualisys Arqus motion capture system(10ea)

setup This system included one tag and eight anchors configured in a rectangular cuboid formation.

If you want to test in your environment, you need to change the config file(config/params.yaml).

cd catkin_workspace/src    
git clone --recursive https://github.com/nooploop-dev/nlink_parser.git 
git clone https://github.com/KYH04444/Tightly_Coupled_Fusion.git
cd ..   
catkin_make

roslaunch tightly_coupled all.launch

Experimental results

tightly_efk_esekf tightlycoupled_test2
tightly_position tightly_orien

Position RMSE (meter)

Filter X RMSE Y RMSE Z RMSE
EKF 0.1573 0.1211 0.1479
ESKF 0.1573 0.1212 0.1480
LIEKF 0.1567 0.1204 0.1513
UKF 0.2446 0.2083 0.1285

Orientation RMSE (degree)

Filter Roll RMSE Pitch RMSE Yaw RMSE
EKF 9.2247 8.8638 33.5447
ESKF 9.2231 8.8654 15.8335
LIEKF 8.5259 9.2223 9.7865
UKF 9.5407 9.3895 17.0105