Lieu : AIP-PRIMECA – Site INSPÉ
Durée : 5 à 6 mois
Début au plus tôt :  février 2022 – Fin au plus tard : 22 juillet 2022 (ou septembre )
Gratification : Oui
Contacts : Vincent Bonnet <vincent.bonnet@laas.fr>, Michel Taix <michel.taix@laas.fr>
Niveau : Master 2 ou PFE d’écoles d’ingénieurs

Drilling is a very important part of the assembly of an aircraft as it involves ten of thousand holes for a single plane.
It is usually accomplished manually by an operator that uses a hand drill to drill a hole. Even if it is highly flexible,
manual assembly suffers in terms of quality, repeatability, and speed. Moreover, musculo-squelettal disorders are
observed for the operators that are performing such difficult tasks. Thus, aircraft manufacturers are nowadays trying
to develop a collaborative approach between a human and a robot to improve task efficiency while keeping the
skills and the flexibility brought by the operator. The spread of new collaborative robotics applications is highly linked to the improvement of human-robot physical
interactions in unknown environments. For those, an efficient dynamics controller, force or torque controller, is of
crucial importance [1]. Dynamics controllers are based on the dynamics model of the robot. To perform correctly,
such controllers require an accurate estimate of Segments Inertial Parameters (SIP). SIP refers to the mass, the 3D
center of mass (CoM) and the six elements of the inertia matrix of each segment of a mechanical structure.
In robotics, SIP are usually extracted from CAD data. However, CAD data do not take into account cabling, covers,
glued components, or the several, sometimes daily, modifications of the end-effector of a multipurpose manipulator.
Therefore during the last decades, several groups have developed methods to identify the dynamics of all kinds of
robots [2].
Once the SIP are identified the new parameters can be used in a dynamics controller such as the computed-torque
method and/or in simulation for planning an interaction task. In this context, we would like to work toward a
collaborative demonstration platform for S.MART AIP-PRIMECA Occitanie. Specifically, the steps of this project are:

  • To create the dynamics model and the simulation environment of the collaborative robot Yaskawa using
    Pinocchio library (https://github.com/stack-of-tasks/pinocchio)
  • To implement in Python an identification pipeline to estimate the SIP of the robot and of its end-effector. To
    perform the experiments on the real robot.
  • To implement in ROS a computed torque controller and an impedance controller [3] and to assess their
    performance for a simple drilling task.

Desired experiences or qualifications:

  • Good skills in Python/C++/ROS
  • Robotics kinematics and dynamics modeling
  • Linear algebra and basic knowledge in optimization
  • Taste for experiments
  • Ability to work in English as the project will in part be supervised by a Phd student speaking English.

References:
[1] M.H. Raibert, J.J. Craig, Hybrid position/force control of manipulators, ASME J. Dyn. Syst. Meas. Control 103
(1981) 126–133.
[2] M. Gautier, W. Khalil, On the identification of inertial parameters of robots, in: IEEE Int. Conf. on Decision
Control, 1998, pp. 2264–2269.
[3] W. Khalil, E. Dombre, Identification of the dynamic parameters, in: Modeling,
Identification and Control of Robots, first ed., Butter-worth-Heinemann, UK, 2002, pp. 291–311.

Interested candidates can send their CV together with a motivation letter to:
Vincent Bonnet <vincent.bonnet@laas.fr>, Michel Taix <michel.taix@laas.fr>