cheetah_inekf_realtime

This project contains a contact-aided invariant-ekf for MiniCheetah. It directly takes the contact estimation results from our deep-contact-estimator via LCM, and output the estimated pose in ROS.

This repository has the following features:

• Directly compatible with the deep-contact-estimator.
• Use ROS for easier communication with other robotics program.
• Tested real-time performance on a Jetson AGX Xavier along with the deep contact estimator.

Dependencies

• lcm 1.4.0
• invariant-ekf/devel Note: Remember to checkout to the devel branch.
• Eigen3
• Boost
• YAML

Configuration

Parameters can be modified in config/settings.yaml:

• project_root_dir: Filepath to your installation directory for this repo
• estimator_enable_debug: Enable debug print on the screen.
• estimator_publish_visualization_markers: Enable LCM publisher for the estimated pose.
• estimator_lcm_pose_channel: Name of the LCM channel for output robot pose.
• estimator_static_bias_initialization: Enable static bias initialization using the first several measurements from IMU.
• system_enable_pose_publisher: Enable pose logger and publish the robot pose over ROS. Enable this will write down the estimated pose in a txt file and publish the pose to ROS at the same time.
• system_inekf_pose_filename: Path for the logged txt file. Kitti means the file will be recorded following the Kitti format.
• system_inekf_tum_pose_filename: Path for the logged txt file in TUM format.

Helpful Commands:

Generating LCM Types:

1. cd cheetah_inekf_lcm_root_directory/scripts
2. bash ./make_types.sh

Running Cheetah Estimator

1. cd ~/${PATH_TO}/catkin_ws
2. In a new terminal in the catkin_ws, do catkin_make (perhaps multiple times)
3. Run source ~/devel/setup.bash
4. roslaunch cheetah_inekf_lcm cheetah_estimator
5. Run lcm-logplayer-gui NAME_OF_LCM_LOG_FILE_HERE
6. The terminal should begin printing out the robot state if the settings.yaml output variables are enabled

Visualizing InEKF in Rviz

1. Start running the cheetah estimator using the instructions above
2. Enter rviz in the terminal
3. Select Add by topic setting and select path
4. Changed fixed frame to the same value as map_frame_id in config/settings.yaml
5. The robot pose should begin being drawn in rviz

Citation

If you find this work useful, please kindly cite our publication in 2021 Conference on Robot Learning:

• Tzu-Yuan Lin, Ray Zhang, Justin Yu, and Maani Ghaffari. "Legged Robot State Estimation using Invariant Kalman Filtering and Learned Contact Events." In Conference on robot learning. PMLR, 2021

@inproceedings{
   lin2021legged,
   title={Legged Robot State Estimation using Invariant Kalman Filtering and Learned Contact Events},
   author={Tzu-Yuan Lin and Ray Zhang and Justin Yu and Maani Ghaffari},
   booktitle={5th Annual Conference on Robot Learning },
   year={2021},
   url={https://openreview.net/forum?id=yt3tDB67lc5}
}