manifold_calibration

ROS package for calibrating a differential drive mobile base with a Lidar sensor for the mk-ros project: https://german-m-garcia.gitbook.io/mk-ros/

Recording data

Your robot will normally have a static transformation being published for the Lidar frame of reference (e.g. "laser") with respect to the "base_link" frame. For the calibration routine to work this tf should not be published. Instead, a scan matcher should publish the transformation of the "laser" frame with respect to an odometry frame of reference, e.g. "odom". Additionally to the tf topic, you should record the encoder tics for each wheel. If you want to visualize the data you can also record the laser scans and other static tfs that you might have:

rosbag record tf tf_static scan /makeblock/lwheel /makeblock/rwheel 

Launching the calibration node

A launch file is provided in the launch directory that uses the bag file provided in the data directory. The parameters to be set are the following:

Parameter	Type	Desc
left_encoder_topic	string	ROS topic for the left wheel encoder tics
right_encoder_topic	string	ROS topic for the right wheel encoder tics
iterations	int	Number of iterations for the solver
radius_left	double	Left wheel's radius in m
radius_right	double	Right wheel's radius in m
tics_per_revolution	int	Encoder tics per wheel revolution
baseline	double	Distance between wheel axes
sensor_x	double	2D Pose of the Lidar sensor. X offset in m
sensor_y	double	2D Pose of the Lidar sensor. Y offset in m
sensor_theta	double	2D Pose of the Lidar sensor. Angle in radians

Publications

This is an implementation of the method described in
Maurilio Di Cicco, Bartolomeo Della Corte and Giorgio Grisetti. "Unsupervised calibration of wheeled mobile platforms" In Proc. of the IEEE International Conference on Robotics and Automation (ICRA), Stockholm, 2016.