[robotics-worldwide] [jobs] PhD opening in Visual/Force Control for Advanced Robotic Manipulation

Paolo Robuffo Giordano paolo.robuffo_giordano at irisa.fr
Wed Mar 14 01:37:08 PDT 2018

The Rainbow team at Inria/IRISA Rennes, France


has an opening for a PhD position in Visual/Force Control for Advanced 
Robotic Manipulation.

*********** DESCRIPTION OF THE ACTIVITIES ***********

Force and vision are the typical sensing modalities for realizing modern 
robot manipulating systems. Force/torque sensing allows fine control of 
the forces exchanged between the robot and the environment (including 
humans), and is a key technology in the recent soft robotics revolution. 
Vision (in all its forms, e.g., mono/stereo, 2D/3D) is needed for 
perceiving the external world and coping with all the unavoidable 
uncertainties of unstructured environments. Research of force/torque 
control and in visual control (also known as visual servoing) for robot 
manipulator has spanned the last decades, albeit quite often these two 
sensing modalities (and associated control problems) are addressed 
separately. This is probably due to the very different characteristics 
of the two sensors: torque/force sensors usually provide very high rate 
measurements that require little processing/interpretation. Vision 
sensors, on the other hand, run at much slower rates and their 
measurements require high levels of processing for being usable by 
control algorithms.

The goal of this PhD is to study how to optimally combine vision and 
force/torque sensing into a common control framework for advanced 
manipulation tasks with a robot manipulator. In particular, our goal is 
to extend the well-known sensor-based control framework for robot 
manipulators (of which the classical visual servoing is a typical 
example) to directly take into account force/torque sensing. Possible 
research directions are:

- how to express and model robotics tasks directly in the sensor space: 
most works on vision/force coupling exploit vision for reconstructing 
the 3D pose of an object of interest (to be grasped), and then apply 
standard force/torque control algorithms for motion and interaction 
control. However, as typical in visual servoing, 3D pose reconstruction 
is often not needed, and one could directly reason in the sensor space 
without the explicit need of solving an inverse (3D reconstruction) 
problem which is poorly robust to miscalibrations

- the very different update rates of vision and force sensing (up to two 
orders of magnitude) motivates the development of advanced 
estimation/control schemes for coping (at runtime) with this 
discrepancy. For instance, model-based filtering can be used for 
interpolating at high rates between consecutive vision measurements, and 
predictive schemes (as in MPC) can be used for anticipating the 
robot/environment behavior when no measurement is available;

- when approaching an object to be grasped, the end-effector will 
typically occlude part of the object and prevent the extraction of the 
needed visual features. Indeed, in many cases the last 
approaching/grasping phase is executed blindly without using the vision 
sensor. Assuming partial, but not total, occlusions, one can still 
exploit modern model-based tracking schemes for still recovering a good 
estimation of the object pose (possibly exploiting a known object 
model). This would allow exploiting vision during all phases, and not 
only when far from the object;

The control algorithms developed during the Ph.D. will be validated on 
the Franka robot available in our lab. This robot has 7 dofs and is 
equipped with torque sensing in all its joints.

*********** CANDIDATE'S EXPECTED PROFILE ***********

The candidate must be a proficient user of C/C++ and ROS. Familiarity 
with matlab/simulink and/or relevant computer vision library (e.g., 
ViSP, OpenCV, PCL) are a plus. Scientific curiosity, large autonomy and 
ability to work independently are also expected.

A M.Sc. degree in computer science, robotics, engineering, applied 
mathematics (or related fields) is required.

*********** SALARY ***********

The position is full-time for 3 years and will be paid according to the 
French salary regulations for PhD students.

***********  ENVIRONMENT ***********

The Rainbow team


is internationally recognized for its scientific activity as well as for 
technology transfer experiences in the field of visual tracking, visual 
servoing, computer vision and sensor-based control for robotics 
applications. The facilities available in the group include three 6-dof 
industrial manipulator arms, a 6-dof torque-controlled backdrivable arm, 
a 7-dof torque-controlled robot arm (Franka), a pioneer indoor mobile 
robot, two humanoid robots (Romeo and Pepper), a fleet of quadrotor 
UAVs, and an indoor testing arena instrumented with Vicon.

The candidates will be under the supervision of

*Dr. Francois Chaumette*


and *Dr. Paolo Robuffo Giordano*.


and will work in close collaboration with other members of the Rainbow 
group involved in the project.

The Rainbow group is part of the Inria/Irisa lab that spreads its 
activities in 30 research teams working in computer science, signal 
processing, and control. It involves about 650 people, including 120 
professors and assistant professors, 100 full-time researchers, 80 
administrative staff, and 250 PhD students.

***********  HOW TO APPLY  ***********

Interested candidates are requested to apply via this form


The position will remain open until a satisfactory candidate is found

Dr. Paolo Robuffo Giordano

Centre National de la Recherche Scientifique (CNRS)
Rainbow Team, Irisa and Inria Bretagne Atlantique
Campus Universitaire de Beaulieu
35042 Rennes cedex, France
Tel: +33 (0)2 99842545
Fax: +33 (0)2 99847171
Email: prg at irisa.fr
Website: https://urldefense.proofpoint.com/v2/url?u=https-3A__team.inria.fr_rainbow_team_prg_&d=DwICaQ&c=clK7kQUTWtAVEOVIgvi0NU5BOUHhpN0H8p7CSfnc_gI&r=0w3solp5fswiyWF2RL6rSs8MCeFamFEPafDTOhgTfYI&m=5CpFWWz4GnuuOqoHmjlBWyywoKei5_5N8U-YjrgEqjM&s=Z7bWqSv7vQUX4XeA0HRT_23xJqtr4nEkWhysh9z1bzM&e=
YouTube: https://urldefense.proofpoint.com/v2/url?u=https-3A__www.youtube.com_channel_UCIHvW3xxlBgXr4-2D9zALju8w&d=DwICaQ&c=clK7kQUTWtAVEOVIgvi0NU5BOUHhpN0H8p7CSfnc_gI&r=0w3solp5fswiyWF2RL6rSs8MCeFamFEPafDTOhgTfYI&m=5CpFWWz4GnuuOqoHmjlBWyywoKei5_5N8U-YjrgEqjM&s=u5suhg0XFPIuFElqUBVm77xBSPI8uvyGJzibaSSM0Eg&e=

More information about the robotics-worldwide mailing list