EN FR
EN FR




Section: New Results

Software development

MECHE toolbox

Participants : Florence Bertails-Descoubes, Gilles Daviet.

The main tool developed in 2011 in the MECHE software was the hybrid iterative solver for Coulomb friction, published in [21] . In 2011, the MECHE software was extensively used to validate this new solver on large data consisting of thousands interacting fibers (subject to tens of thousands frictional contacts). Code parallelization and optimization were performed so as to speed up computations.

Platform development: Siconos

Participants : Vincent Acary, Olivier Bonnefon, Maurice Brémond, Franck Pérignon.

The main achievements for the Siconos platform are

  1. Automatic serialization of the whole set of classes in Siconos

  2. Improvements and development of a full auto-generated Python wrapper in the Siconos/Front-End

  3. Development of the Siconos/Multi-body library and validation on industrial examples (C60 circuit breaker of Schneider Electric)

  4. New algorithms for the resolution of the discrete frictional contact problem

  5. Development of (θ/γ)-schemes for first order dynamical systems

  6. Development of routines for sliding mode control

AMELIF framework

Participants : Pierre-Brice Wieber, François Keith, Jory Lafaye.

The main improvements to the AMELIF framework developed this year are:

  • A new package specific to torque control has been developed, that contains the algorithms required to realize a given motion with a humanoid robot: estimation of contact forces and torque computation (feedforward), feedback methods ensuring the contact force convergence. These algorithms have been tested for two humanoid platforms: the robot Romeo and the robot HRP-2.

  • The dynamics algorithm has been improved, based on the expertise coming from the HuMAnS toolbox. Besides, inverse dynamic algorithms and Runge Kutta integration methods have been added.

  • Finally, the bridge with the stack-of-tasks framework  [55] , that computes the inverse kinematics and the inverse dynamics of humanoid systems, has been enhanced to handle the binding with Python. With this framework, it is possible to use the Model Predictive Control algorithm aforementioned and to simulate the behaviour of a humanoid in a dynamic simulation realized by AMELIF. Tests are still in progress.