This package contains tools for creating, fitting, and tracking models of rigid kinematic tree structures, including human limbs and articulated objects like furniture, boxes, etc. It can:

• Fit the parameters of a rigid kinematic tree model to a specific person/object using motion capture data
• Use a UKF to track the state of a kinematic model in real time using motion capture data
• Stream the real time state estimate of a model on a ROS topic
• Save the identified kinematic tree models to JSON files for later editing and use

Particularly useful files in this repo include:

• src/kinmodel/kinmodel.py: Core library for defining, saving, fitting, and tracking kinematic models
• src/fit_kinmodel.py: Fit a kinematic model's joint parameters to a pre-recorded mocap sequence
• src/track_kinmodel.py: Track a fitted kinematic model online and output estimated joint angles to a ROS topic
• src/track_kinmodel_offline.py: Run kinematic model tracking on a pre-recorded mocap sequence and plot the resulting joint trajectory
• src/mocap_generate_marker_assignments.py: RViz GUI for generating a mapping between mocap marker indices and named features attached to a kinematic model
• src/kinmodel/mocap_recorder.py: Records mocap sequences for later fitting of kinematic models

This package depends on the phasespace ROS package to interface with the mocap systems we have in the TeleImmersion Lab.