ROS_BASALT

This is a ROS wrapper for the Basalt project. It's state-of-the-art visual inertial odometry with open source code release.

This wrapper includes the newest code release from the Basalt project including:

• Square Root Marginalization for Sliding-Window Bundle Adjustment, N. Demmel, D. Schubert, C. Sommer, D. Cremers, V. Usenko, In 2021 International Conference on Computer Vision (ICCV), [arXiv:2109.02182]

---

Attention

Please refer to the Basalt project for details of code implementation and relevant documentations and publications.

This wrapper is made specifically to combine Basalt VIO with Intel REALSENSE T265 while publishing to ROS. Therefore, only a portion of the scripts in the original project are included(performance testing on datasets, testing modules and etc are not included)

---

Tested environment

ROS Noetic on Ubuntu 20.04.4 LTS with Intel REALSENSE T265

---

Installation

1. Clone this package to your catkin workspace src folder

2. Install the libraries in the thirdparty folder if you don't already have them in your enviroment:

   cd to_the_package_not_yet_installed
   

   mkdir build 
   

   cd build
   

   cmake ..
   

   sudo make install
   

3. Install librealsense following this link

4. Build the ros package

   catkin_make --pkg ros_basalt
   

5. Launch

   roslaunch ros_basalt rs_t265_vio.launch
   

   a. Raw IMU readings are published to rs_t265_imu. Note that pose estimation by T265 itself is disabled

   b. Camera images are published to rs_t265/cam0 and rs_t265/cam1

   c. Pose estimation by Basalt VIO is published to basalt/pose

   d. Linear velocity estimation by Basalt VIO is published to basalt/velocity