GitLab

Improve TaskSE3Equality: apply mask before returning task acceleration, desired task acceleration, position error and velocity error.

Extend InverseDynamicsFormulationAccForce to work also for robot manipulators, which are fully-actuated, so they do not need to specify the unactuated base dynamics as equality constraints in the TSID optimization problem.

Extend RobotWrapper to fixed-base robots: if no floating-base root joint is specified, assume robot is fully actuated. Add method na to get number of actuators. Add methods to get frame vel/acc in local world-oriented frame (useful for plotting).

Add computation of average computation time in ex_3, slightly change configuration parameters for tests with Romeo.

Revert "[CMake] disable catkin for now"

This reverts commit 50144111.

[bindings][formulations] add binding for removeFromHqpData

[robotwrapper] create constructor from pinocchio::model

[Notebooks] Fix the waist task - Working examples

The plot_utils.py file should be placed in the parent directory of the notebooks 

Adding and fixing the waist task for Pyrène.
The control is now working on both exercises.
Be careful to use joints and not frames when defining the Tasks and Trajectories. 
ie: use 'root_joint' and not 'base_link' for the waist.

Change the weights and gains: use the one of Pierre Fenrbach.

Convert the exercizes into jupyter notebooks

Change the create_empty_figure() method:
- Add an argument figsize to the method (to (7,7) by default)
- Add the figsize argument to the subplots

Try to add a task on the waist of Pyrène to fix the control (ex1). Not yet working.

Change a little plot_utils to fix the figsize.
Add an help for the installation process which will be added in the wiki.

Add TrajSample.pos(SE3) for python bindings.

Remove duplicate useLocalFrame definition. Fixes regression from #39

Expose the setMask and useLocalFrame for the TaskSE3Equality python

Avoid delegation costructor in Contact6d as it is a C++11 feature.

Expose TaskSE3Equality::useLocalFrame to python

Fix bug in python bindings of InverseDynamicsFormulationAccForce. The methods getContactForces and checkContact were assuming that the contact force vectors were always 12d. This was the case a while ago, but now TSID supports also 3d contact points, so this assumption does no longer hold in general. To fix the bug I have introduced a new method that allows the user to get the contact force vector without knowing its size in advance.

Fix small bug showing up only in DEBUG due to malloc being forbidden before potential resize of QP matrices in EiQuadProgFast

Fix small bug showing up only in DEBUG mode for an inconsistency in size of the Kp, Kd vectors in ContactPoint

Release of version 1.2.0.

as it prevents to make a release

Pinocchio v2, fix #31