Contents
What does vi_sfm do?
vi_sfm is a ROS package used to solve visual-inertial structure from motion problems. This package fuses visual and inertial data in a closed-form solution. The closed-form solution consists of a single equation that determines the initial velocity, the distance to point-features and the gravity linearly by only using inertial measurements unit (IMU) and monocular camera features acquired in a short time interval (~4 seconds). It is assumed that the camera is calibrated and the transformation between the camera and the IMU is known and can be determinated using TF transformation.
Paper
The closed-form solution was presented at ICRA 2016.
Supported hardware
vi_sfm will work with any camera driver node satisfying the standard ROS camera interface and any imu driver node satisfying the sensor
Usage of Robot State
To run the robot_state node :
rosrun vi_sfm robot_state imu:=/my_imu feature:=/my_feature camera:=/my_camera
To run the robot_state node using rectified features :
rosrun vi_sfm robot_state imu:=/my_imu feature:=/my_feature camera:=/my_camera isRectified:=True
Or use a specific launch file :
roslaunch vi_sfm robot_state.launch
Nodes
robot_state
robot_state subscribes to ROS imu, features image and camera info topics. It performs the closed-form solution after 4 seconds of collecting datas.Subscribed Topics
feature (vi_sfm/Features)- image features topic. This message lists a set of features that can be tracked during the time interval.
- imu topic
- camera info topic for camera calibration parameters
robot_state_with_bias
robot_state_with_bias is exactly the same node that robot_state, but used libLBFGS to perform a gradient-descent that can estimated the gyroscope bias.Subscribed Topics
feature (vi_sfm/Features)- image features topic. This message lists a set of features that can be tracked during the time interval.
- imu topic
- camera info topic for camera calibration parameters
