ROS package for robust odometry estimation based on RGBD data
This package implements a viual odometry sysem based on RGBD cameras. The system makes use of the visual+depth RGBD images to robustly estimate the robot motion.
Computer efficiency has been considered in the source code development, it runs smoothly in a single core i7. The algorithms implements the following methods:
- Odometry based on key-framing to reduce the impact of the drift.
- Perspective n Point (PnP) for frame-to-keyframe transform estimation.
- Optional loose coupled integration of IMU roll and pitch angles.
- Window bundle adjustment for pose refinement (not map).
The source code has been tested in ROS indigo with Ubuntu Lunux 14.04. However, no major requirements are needed except the software packages listed in Dependencies
This package depends on non-linear optimization ROS package than can be downloaded here (https://github.com/robotics-upo/nonlinear_optimization)
In order to build the package, git clone this package (and the nonlinear_optimization) into your src directory of your Catkin workspace and compile it by using catkin_make as usual.