roboptim-retargeting

Overview

RobOptim toolbox aiming at processing Motion Capture Data and in particular retarget motion capture data in order to allow a particular robot to replace the motion.

This toolbox relies on the following software:

• Choreonoid (rigid body computation)
• libmocap (motion capture data loading)

Optionally, if metapod is available, some constraints have also been implemented using this library.

It provides three command-line tools which are in fact the three steps of the conversion process:

• roboptim-retargeting-markers
• roboptim-retargeting-marker-to-joint-converter
• roboptim-retargeting-joints

The marker optimization optimizes the markers positions in the Euclidian space. If the motion capture matches a human skeleton, this step will adapt the markers positions at each frame to make it match the robot morphology.

Marker to Joint converter converts marker positions to joints trajectories of a particular robot.

The joint optimization will, at least, reshape the joint trajectories to satisfy robotics constraints such as balance (ZMP), actuators limits (position, velocity, torque).

RobOptim Functions

All functions are placed into the include/roboptim/retargeting/function folder.

Acceleration (joint-based optimization and marker-based optimization)

#include <roboptim/retargeting/function/acceleration.hh

Use StateFunction to compute each joint/marker acceleration over three frames. This function can be used to penalize large motion as a cost function.

Marker Laplacian Deformation Energy (marker-based optimization)

#include <roboptim/retargeting/function/marker-laplacian-deformation-energy/choreonoid.hh>

Compute the Marker Laplacian Deformation Energy for markers.

Body Laplacian Deformation Energy (joint-based optimization)

#include <roboptim/retargeting/function/body-laplacian-deformation-energy/choreonoid.hh>

Compute the Marker Laplacian Deformation Energy for bodies.

Bone Length (marker-based optimization)

#include <roboptim/retargeting/function/bone-length/choreonoid.hh>

Compute bone length for each segment and compare it to the robot segments.

Joints limits (joint-based optimization)

#include <roboptim/retargeting/function/joint-limits/choreonoid.hh>

Compute joints positions and/or velocities and make sure they are in the actuators limits.

Joints torque limits (joint-based optimization)

#include <roboptim/retargeting/function/torque/choreonoid.hh>

Compute joints torques and make sure they are in the actuators limits.

Body position (joint-based optimization)

#include <roboptim/retargeting/function/forward-geometry/choreonoid.hh>

Force a particular body to stay at the correct location. Typically, it is used to force the feet on the floor.

ZMP (joint-based optimization)

#include <roboptim/retargeting/function/zmp/choreonoid.hh>

Compute the Zero-Momentum Point attached with the robot. If it stays in the support-polygon then the motion will remain balanced.

RobOptim Problems

Marker Optimization

Cost: Laplacian Deformation Energy (marker-based) and optionally acceleration.

Constraints:

• Bone Length (linear function)

Joint Optimization

Cost: Laplacian Deformation Energy (joint-based) and optionally acceleration.

Constraints:

• Joint Limits (position, velocity)
• Body Position (feet on the ground)
• Torque
• ZMP