The site is read-only. Please transition to use Robotics Stack Exchange
| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
 | 1 | initial version |
To answer the question of how it works, I will provide some sources to look into. The Kalman filter when used for localization is typical in robotics.
Textbook with chapter on ekf localization
Also, if you have access to a university library, the following book is quite good: Principles of Robotic Motion: Theory, Algorithms, and Implementation. MIT Press, Cambridge, MA, 2004.
Since the robot was build and provided ROS ready, the kinematic model and measurement model is likely already tuned by Clearpath robotics. I doubt you would want to change anything there, although I'm not sure how to actually gets your hands on it. Perhaps an email to Clearpath could be an option if nothing is mentioned in the documentation or user manuals.
Regarding the bias question: I'm not familiar with the ekf node in ROS but EKFs in general fuse data based on the measurement model of the system. If one sensor is more noisy (higher covariance) than the other, then the estimation of the robot state will be biased towards using the information from the better sensor. Again, since Clearpath built the robot, they likely already tuned the measurement model based on the hardware specs of the IMU and encoders.
Unfortunately I couldn't give you any direct answers but I hope this helps.
ROS Answers is licensed under Creative Commons Attribution 3.0 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.
