Courses related from third semester programme (indicate the selected courses by “X”):

Advanced modeling of Robots, X Optimal design of Robots, X Humanoid Robots, Identification and Control of Robots, Bio-Robotics, Vision based control of Robots, Acquisition and Simulation of Human motion.

Motivations and general objectives:

This thesis is part of a project named ARROW funded by the French Research National Agency. ARROW aims at designing and controlling accurate and high-speed robots for pick-and-place and laser cutting operations.

Parallel robots are now more and more used for high-speed pick-and-place operations. As a consequence, we have recently synthesized an exhaustive list of two-dof translational parallel manipulators This thesis subject aims to analyze the stiffness of those manipulators and determine an (some) intrinsic stiffness index (indices) that could be used at the conceptual design stage.

Proposed work plan:

First, the candidate will have to understand the type-synthesis of two-dof translational parallel manipulators carried out in the master thesis of Latifah NURAHMI [1] and double check the results. Figure 1 illustrates a two-dof translational parallel manipulator, named IRSBot-2, synthesized in the scope of ARROW project and studied in [2]. Then, the stiffness of twenty manipulators selected from the exhaustive list of synthesized manipulators will be analyzed and compared carefully. The dimensions of those manipulators should be optimized with regard to a given regular workspace to reach and an accuracy to respect.Finally, an (some) intrinsic stiffness index (indices) should be developed to compare parallel manipulators at their conceptual design stage with regard to their stiffness. Some intrinsic stiffness indices exist in the literature, but are not relevant [3]. Note that only information about the type and number of joints, relative orientation of neighboring joints, number of loops in case of parallel robot, type and diversity of actuators are given at the robot conceptual design stage [4], [5].

Figure 1: CAD Modeling of the IRSBot-2

Motivations and general objectives:

Dynamic control of high speed robots like parallel robot requires high speed estimation of the state of the robot. The control can be done in joint space, in cartesian space and more generally in sensor space. Model Based Dynamic Control approaches use proprioceptive sensors (i.e. joint sensors). Dynamic Sensor Based Control approaches use exteroceptive sensors (i.e cameras).

In 2006, [Aider06] has shown that it is possible to estimate both Pose and Velocity (12D) state using a Rolling shutter camera. In 2009, [AIDER09] extends this work for a stereo rig of rolling shutter cameras. More recently this work has been applied to Random access CMOS camera. The control has been developed for three different robots (Orthoglide, T3R1 and Quattro).

The work of the master will focused on the estimation of Pose and velocity using one and two camera using different assumptions for the motion model. Algorithms will be dedicated to new high speed CMOS camera. Evaluation will done using the parallel robot orthoglide. Algorithms will be developed under C++ and optimize in order to reach estimation near 1 or 2kHz.

In case the work will have a very quick development and evaluation, we will try to develop new algorithms able to estimate the direction of the legs and its first derivative at the same time.

Proposed work plan:

Bibliography analysis

Pose/Velocity estimation algorithm in C++ for random access CMOS camera

Evaluation of the performance (comparison with ground truth)

Real implementation on orthoglide

Possible extension

Motivations and general objectives:

This thesis concerns the identification and control of flexible robots. The identification of the elastodynamic model of flexible robot is an actual challenge for researchers. This model is of great importance for the accurate end-effector motion control of flexible robots. Several works have dealt with the model identification using external measures of the link displacements (accelerometers, laser trackers, etc.) or actuator torques. However, to the best of our knowledge, the identification of the flexible dynamic parameters of robots using measures of the base reactions has never been carried out. Such work is of interest as force plates are much less sensitive to noise than the other measurement systems. This master’s work aims at filling this gap.

Proposed work plan:

The work plan is the following:

• Modeling of a flexible 1 DOF serial robot composed of a linear actuator (EMPS) on which is added a flexible beam;
• Design of exciting trajectories for identification;
• Measures on the experimental setup;
• Identification of the dynamic parameters of the flexible robot, using either the base reaction forces or the input torques.
• Design of the controller for obtaining accurate end-effector motion