ROS package for mapping a motion capture trajectory to a robot joint space - gradient based

This C++ ROS package utilizes the Sparse Nonlinear OPTimizer (SNOPT) software package to determine an optimal mapping between the motion capture lab reference frame and the robot joint space. This mapping preserves the kinetic properties of the trajectory while minimizing the robot joint usage. The related identifiers also provide a C++ ROS package for performing the aforementioned mapping, but which utilizes the constrained optimization by linear approximation (COBYLA) numerical optimization method. In practice, the COBYLA-based optimization method outperforms the SNOPT-based optimization when the initial input guess to the optimization algorithm is far from optimal. However, once a close to optimal joint-space trajectory has been found, the SNOPT-based optimization outperforms the COBYLA-based optimization for minimizing joint velocity usage while satisfying joint and joint velocity limits.

Also see related identifiers for the Motion Capture to Robot software package, which provides a full suite of algorithms for robotically replicating motion capture datasets. The related identifiers also provide a dataset that supports this package as well as the Motion Capture to Robot software package.

The user manual PDF is attached to the release. Example configuration files for the optimization algorithm are also attached. Manual.pdf configurationFiles.zip

Related publication can be found at: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0242005