Pose Prediction for Mobile Ground Robots

Hello everyone,

I want to announce the open source release of sdf_contact_estimation, a ROS1 noetic library for the fast and accurate pose prediction of mobile ground robots in rough terrain (also referred to as robot settling or robot-terrain interaction). A demo is available.

With a runtime of about 0.5 ms per predicted pose, this approach is much faster than a traditional physics simulation and suitable for online planning applications. Voxblox is used as the environment model. Possible use-cases are:

  • Automatic flipper control
  • Path planning
  • Tip-over prevention

This software has been used by Team Hector in the RoboCup Rescue League for whole-body planning with a tracked robot with actuated flippers. That software will be part of a future release.

License: MIT

I am looking forward to your feedback.

Best wishes,
Martin Oehler

7 Likes

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.