- OpenSim (tested with OpenSim 4.3)
- Hi-ROS skeleton_msgs
Before running the ROS node OpenSim must be patched.
Add the following code in OpenSim InverseKinematicsSolver.h:
- after line 234:
inline void updateMarkersReference(std::shared_ptr<MarkersReference> newMarkersReference)
{
_markersReference = newMarkersReference;
}
inline void updateOrientationsReference(std::shared_ptr<OrientationsReference> newOrientationsReference)
{
_orientationsReference = newOrientationsReference;
}
inline void updateMarkersWeight(const SimTK::Real& newMarkersWeight)
{
_markers_weight = newMarkersWeight;
}
inline void updateOrientationsWeight(const SimTK::Real& newOrientationsWeight)
{
_orientations_weight = newOrientationsWeight;
}
- after line 269:
// Markers/orientations relative weights
SimTK::Real _markers_weight{1.};
SimTK::Real _orientations_weight{1.};
Edit the following lines in OpenSim InverseKinematicsSolver.cpp:
updAssembler().adoptAssemblyGoal(condOwner.release(), _markers_weight);
updAssembler().adoptAssemblyGoal(condOwner.release(), _orientations_weight);
Skeleton's marker names must correspond to the virtual markers defined in the model. Skeleton's link names must correspond to the virtual IMUs defined in the model.
ros2 launch hiros_opensim_ik_solver default.launch.py
Please cite the following paper:
M. Guidolin, M. Vanuzzo, S. Michieletto and M. Reggiani, "Enhancing Real-Time Body Pose Estimation in Occluded Environments Through Multimodal Musculoskeletal Modeling," in IEEE Robotics and Automation Letters, vol. 9, no. 12, pp. 10748-10755.
Bib citation source:
@ARTICLE{10714023,
author={Guidolin, Mattia and Vanuzzo, Michael and Michieletto, Stefano and Reggiani, Monica},
journal={IEEE Robotics and Automation Letters},
title={Enhancing Real-Time Body Pose Estimation in Occluded Environments Through Multimodal Musculoskeletal Modeling},
year={2024},
volume={9},
number={12},
pages={10748-10755},
keywords={Real-time systems;Accuracy;Optimization;Biological system modeling;Pose estimation;Cameras;Robot sensing systems;Safety;Biomechanics;Tracking;Human detection and tracking;human-robot collaboration;multi-modal perception for HRI;RGB-D perception;sensor fusion},
doi={10.1109/LRA.2024.3478569}
}