- David Marquez-Gamez (IRT Jules Verne)
- Sylvain Vandernotte (IRT Jules Verne)
- Inigo Rodriguez (IRT Jules Verne)
- Olivier Kermorgant (IRT Jules Verne)
- Alexis Girin (IRT Jules Verne)
- David Guérin (IRT Jules Verne)
- Lorenzo Gagliardini (IRT Jules Verne)
- Paul Yvain (IRT Jules Verne)
IRT Jules Verne is a research institute developing innovation for the industry in the field of Manufacturing. The Robotics and Cobotics (ROC) team at IRT JV is composed of five full time PhD/engineers and three PhD students. We favoured a constructive and integrative way of thinking robotics, aiming at defining robotics as a complete discipline and a challenging field of work. The sum of skills gathered by the team members covers all today’s wide variety of problems: from mechanical design to high level task programming.
Lorenzo, David G. and Paul are specialized in mechanical design and robot modelling and identification. Alexis, Olivier and Sylvain tackle control problems for robots with fix or mobile base. The latter are well-trained to the task function approach (visual servoing or sensor-based control) which allows encapsulating complex robot behaviour inside control functions. Current work tends to integrate this task formalism mixed with a torque control scheme in order to define force-related task for collaborative robots. David MG. is an expert in perception for autonomous robots. Currently he is working on embedded perception systems to solve several problems such as SLAM with DATMO, semantic mapping, object classification/identification, for a distributed heterogeneous multi robot system. Inigo is specialized in mechatronics, being able, for example, to write code, propose electronics solutions or design mechanical parts.
One of our strongest interests is to imagine and design solutions able to be implemented as fast as possible at shop floor level. That is why for some projects, it has been decided to use quasi-realistic simulator such as Gazebo software driven by the middleware ROS to design and integrate modular and easy-to-deploy pieces of software.
One of the main goals that our team aims with this challenge is to develop robot capabilities in order to be flexible and to bring added-value to industrial processes.
The robot needs to be able to adapt its motion capabilities to navigate in a dynamic factory in autonomy where only walls are immobile and the surroundings are composed of static objects to cross or avoid, long flat surface with holes, sloping sections and operators in motion or in working situation. Thus, the robot will run in a non-controlled, non-deterministic and complex environment. It is composed of a manipulator arm fixed on a mobile base that makes the system highly redundant.
To achieve this challenge we intend to combine two approaches:
1.) In terms of control, the problem can be divided in several functions in the sense of the task formalism. Current works tend to divide part placement for assembly problem into simpler sub tasks. In our previous work , we simulate a precise positioning for part assembly based on this concept. This can be also applied on an omnidirectional mobile base in order to have a unified way to control all degrees of freedom of the entire robot . The task formalism can be extended to torque controlled robot, which brings the possibility to specify task with force data and then make the manipulation sure either for environment or operators.
2.) The challenge is also about perception and navigation problems. Even if the robot is able to manage any kind of undesired contacts, the better is to avoid them whenever possible. Active SLAM combined with detection and tracking of moving objects, is the other key to solve the problem. In this sense, our team works in this field since several years by means of several fundamental and industrials projects we have completed and others which are currently being executed [3, 4, 5]. Additionally our team also has skills in 3D model fitting and pose estimation of CAD models. This allows us to perform manipulation, detect parts and grasp them . Furthermore 3D model fitting can also be used at a larger scale in order to localize the robot in a plant model, ensuring efficient high level navigation planning and logistics .
 Performing Assembly Tasks Under Constraints Using 3D Sensor-based Control, Vandernotte S., Chriette A., Suarez Roos A. and Martinet P., Proceedings of the 13th International Conference of Intelligent Autonomous System (2014), accepted in april 2014
 ASIMOV (IRT JV National project): design, manufacturing and validation of mobile cobot devoted to assembly task in aeronautic area
 Active Visual-Based Detection and Tracking of Moving Objects from Clustering and Classification Methods, Marquez-Gamez D. and Devy M., Advanced Concepts for Intelligent Vision Systems, p361-373, 2012
 Environment reconstruction and navigation with electric sense based on kalman filter, Lebastard V., Chevallereau C., Girin A., Servagent N., Gossiaux P.-B., Boyer F. International Journal of Robotics Research 32, 2 (2013) 172 - 188 [hal-00794706 - version 1]
 CHARLIE (ANR funded national project proposal): Development of distributed perception and supervision for a cobot fleet
 FAST (IRT JV National project): Development of smart gripper and safe algorithms for industrial arms. Application to automotive industry in collaborative context
 IDEAL COWORKER (H2020 funded European project proposal): Industrial DEpendAble Legged and Collaborative multi-Operation WORKER. An Industry proof flexible cobot for the European manufacturing industry