Recent PhD defenses

Motion planning for elastic rods

By Olivier Roussel
Defended October the 5th, 2015

The motion planning problem has been broadly studied in the case of articulated rigid body systems but so far few work have considered deformable bodies. In particular, elastic rods such as electric cables, hydraulic or pneumatic hoses, appear in many industrial contexts. Due to complex models and high number of degrees of freedom, the extension of motion planning methods to such bodies is a difficult problem. By taking advantage of the properties of static equilibrium configurations, this thesis presents several approaches to the motion planning problem for elastic rods.

Keywords: Motion Planning, Elastic Rods, Optimal Control

Jury: S. Hutchinson (Univ. of Illinois at Urbana-Champaign), F. Boyer (Ecole des Mines de Nantes / IRCCyn), B. Bayle (Télécom Physique Strasbourg / Icube), P. Danès (LAAS-CNRS / Univ. de Toulouse III), P. Souères (LAAS-CNRS), M. Taïx (LAAS-CNRS / Univ. de Toulouse III)

Watch the video of the defense (in French) -- Read the manuscript (in French)

Semiotics of Motion: Toward a Robotics Programing Language

By Nicolas Mansard
Defended Jully the 1st, 2013

My work is aiming at establishing the bases of a semiotics of motion, in order to facilitate the programing of complex robotics systems. The objective is to build a symbolic model of the action, based on the analysis of the numerical functions that drive the motion (control and planning). The methodology comes from the well-known robotics concepts: motion-planning algorithms, control of redundant systems and task-function approach. The originality of the work is to consider the task as the unifying concept both to describe the motion and to control its execution.

The document is organized in two parts. In the first part, the task-function control framework is extended to cover all the possible modalities of the robot. The objective is to absorb from the lowest-possible functional level the maximum of uncertainty factors. It is then not any more necessary to model these factors at the higher functional levels. This sensorimotor layer is then used as a basic action vocabulary that enables the system to be controlled with a higher-level interface. In the second part, this action vocabulary is used to provide a dedicated robotics programing language, to build motion-planning methods and to describe an observed movement.

The proposed methods are generic and can be applied to a various systems, from robotics (redundant robots) to computer animation (virtual avatars). Nonetheless, the work is more specifically dedicated to humanoid robotics. Without forgetting other possible outlets, humanoid robotics provides a tangible applicative and experimental framework. It also leads toward the natural human motion, as presented in the end of the document.

Keywords: Robotic, redundant systems, anthropomorphic mouvement, sensor-based control, task sequencing, obstacle avoidance

Jury: A. De Luca (Univ Sapienza, Roma, Italy), P. Fraisse (LIRMM), E. Todorov (Univ Washington, USA), F. Chaumette (INRIA Bretagne), A. Kheddar (JRL-Japan), J-B. Hiriart-Urruty (UPS, Toulouse), J-P. Laumond (LAAS), B. Duprieu (Airbus)

Planning Optimal Motions for Anthropomorphic Systems

By Antonio El Khoury
Defended June the 3rd, 2013

This thesis deals with the development and study of algorithms for planning optimal motions for anthropomorphic systems, which are underactuated and highly redundant systems, such as humanoid robots and digital actors. Randomized motion planners and optimal control methods are proposed and discussed. A first contribution concerns the use of an efficient graph search algorithm in order to optimize walk trajectories that were previously obtained for a bounding-box representation of the system using randomized motion planners. The second contribution develops the use of constrained randomized motion planners in order to plan in a generic way whole-body motions that involve both walking and manipulation. Finally we develop an algorithmic approach which combines constrained randomized motion planners and optimal control methods; this approach allows the generation of dynamic, fast and collision-free motions for anthropomorphic systems in the presence of obstacles.

Keywords: Motion Planning, Optimal Control, Anthropomorphic System, Humanoid Robot

Jury: B. d'Andréa-Novel (Mines ParisTech), M. Bennewitz (Univ. of Freiburg), T. Bretl (Univ. of Illinois at Urbana-Champaign), P. Danès (LAAS/Univ. de Toulouse III), R. Gelin (Aldebaran Robotics), A. Kheddar (JRL, CNRS/AIST), F. Lamiraux (LAAS/CNRS), Michel Taïx (LAAS/Univ. de Toulouse III)

Watch the video of the defense -- Read the manuscript -- See the defense slides (all in English)

Vision Based Motion Generation for Humanoid Robots

By Olivier Stasse
Defended April the 4th, 2013

This manuscript present my research activities on real-time vision-based behaviors for complex robots such as humanoids. The underlying main scientific question structuring this work is the following: ``What are the decisional processes which make possible for a humanoid robot to generate motion in real-time based upon visual information ?'' In soccer humans can decide to kick a ball while running and when all the other players are constantly moving. Recast as an optimization problem for a humanoid robot, finding a solution for such behavior is generally computationally hard. For instance, the problem of visual search consider in this work is NP-Hard.

The first part is concerned about real-time motion generation. Starting from the general constraints that a humanoid robot has to fulfill to generate a feasible motion, some core problems are presented. From this several contributions allowing a humanoid robot to react to change in the environment are presented. They revolved around walking pattern generation, whole body motion for obstacle avoidance, and real-time foot-step planning in constrained environment.

The second part of this work is concerned about real-time acquisition of knowledge on the environment through computer vision. Two main behaviors are considered: visual-search and visual object model construction. They are consider as a whole taking into account the model of the sensor, the motion cost, the mechanical constraints of the robot, the geometry of the environment as well as the limitation of the vision processes. In addition contributions on coupling Self Localization and Map Building with walking, real-time foot-steps generation based on visual servoing will be presented.

Finally the core technologies developed in the previous contexts were used in different applications: Human-Robot interaction, tele-operation, human behavior analysis. Based upon the feedback of several integrated demonstrators on the humanoid robot HRP-2, the last part of this thesis tries to draw some directions where innovative ideas may break some current technical locks in humanoid robotics.

Keywords: computer vision - control - humanoid robotics - motion planning

Jury: F. Chaumette (Lagadic/INRIA, Rennes), C. Chevellareau (CNRS, IRCyNN Nantes), A. Kheddar (JRL, CNRS/AIST), F. Lerasle (LAAS/Univ. de Toulouse III ), P. Poignet (LIRMM, Montpellier), G. Sandini (IIT, Italy) F. Lamiraux and J-P. Laumond (LAAS/CNRS)

See the video of the defense -- Read the manuscript (both in English)

Manipulation and locomotion for humanoid robotics using real-time footprint optimization

By Duong Dang
Defended October the 30th, 2012

This thesis focuses on realization of tasks with locomotion on humanoid robots. Thanks to their numerous degrees of freedom, humanoid robots possess a very high level of redundancy. On the other hand, humanoids are underactuated in the sense that the position and orientation of the base are not directly controlled by any motor. These two aspects, usually studied separately in manipulation and locomotion research, are unified in a same framework in this thesis and are resolved as one unique problem. Moreover, the generation of a complex movement involving both tasks and footsteps is also improved becomes reactive. By dividing the optimization process into appropriate stages and by feeding directly the intermediate result to a task-based controller, footsteps can be calculated and adapted in real-time to deal with changes in the environment. A perception module is also developed to build a closed perception-decision-action loop. This architecture combining motion planning and reactivity validated on the HRP-2 robot. Two classes of experiments are carried out. In one case the robot has to grasp an object far away at different height level. In the other, the robot has to step over an object on the floor. In both cases, the execution conditions are updated in real-time to deal with the dynamics of the environment: changes in position of the target to be caught or of the obstacle to be stepped over.

Keywords: manipulation, locomotion, footstep optimization, realtime, adaptation, computer vision, visual servoing, reactivity

Supervised by J-P. Laumond
Jury: P-Y Oudeyer (Flower/INRIA, Bordeaux), P. Fraisse (LIRMM/Univ Montpellier), F. Lamiraux and J-P. Laumond (LAAS/CNRS)

See the video of the defense -- Read the manuscript (both in French)

Numerical Optimization for robotics and closed-loop trajectory execution

By Thomas Moulard
Defended September the 18th, 2012

The presented work is divided into two parts. In the first one, an unified computer representation for numerical optimization problems is proposed. This model allows to define problems independently from the algorithm used to solve it. This unified model is particularly interesting in robotics where exact solutions are difficult to find. The second part is dealing with complex trajectory execution on humanoid robots with sensor feedback. When a biped robots walks, contact points often slip producing a drift which is necessary to compensate. We propose here a closed-loop control scheme allowing the use of sensor feedback to cancel execution errors. To finish, a method for the the development of complex robotics application is detailed. This thesis contributions have been implemented on the HRP-2 humanoid robot. Supervised by F. Lamiraux
Jury: T. Géraud, D. Filliat, R. Gelin, P. Souères, F. Lamiraux

See the video of the defense -- Read the manuscript -- See the defense slides (all in French)

Task recognition by reverse control

By Sovannara Hak
Defended November the 2nd, 2011

Efficient methods to perform motion recognition have been developed usingstatistical tools. Those methods rely on primitives learning in a suitablespace, for example the latent space of the joint angle and/or adequate taskspaces. The learned primitives are often sequential : a motion is segmentedaccording to the time axis. When working with a humanoid robot, a motion can bedecomposed into simultaneous sub-tasks. For example in a waiter scenario, therobot has to keep some plates horizontal with one of his arms, while placing aplate on the table with its free hand. Recognition can thus not be limited toone task per consecutive segment of time. The method presented in this worktakes advantage of the knowledge of what tasks the robot is able to do and howthe motion is generated from this set of known controllers to perform a reverseengineering of an observed motion. This analysis is intended to recognizesimultaneous tasks that have been used to generate a motion. The method relieson the task-function formalism and the projection operation into the null spaceof a task to decouple the controllers. The approach is successfully applied ona real robot to disambiguate motion in different scenarios where two motionslook similar but have different purposes.

Supervised by N. Mansard, O. Stasse and J-P. Laumond
Jury: A. Billard (EPFL), B. Espiau (INRIA-Grenoble), J-P. Laumond, N. Mansard, R. Alami, O. Stasse (LAAS, CNRS)

Read the manuscript (in French)

Generating whole body movements for dynamic anthropomorphic systems under constraints

By Layale Saab
Defended October the 31th, 2011

This thesis studies the question of whole body motion generation for anthropomorphic systems. Within this work, the problem of modeling and control is considered by addressing the dicult issue of generating human-like motion. First, a dynamic model of the humanoid robot HRP-2 is elaborated based on the recursive Newton-Euler algorithm for spatial vectors. A new dynamic control scheme is then developed adopting a cascade of quadratic programs (QP) optimizing the cost functions and computing the torque control while satisfying equality and inequality constraints. The cascade of the quadratic programs is de ned by a stack of tasks associated to a priority order. Next, we propose a uni ed formulation of the planar contact constraints, and we demonstrate that the proposed method allows taking into account multiple non coplanar contacts and generalizes the common ZMP constraint when only the feet are in contact with the ground.

Then, we link the algorithms of motion generation resulting from robotics to the human motion capture tools by developing an original method of motion generation aiming at the imitation of the human motion. This method is based on the reshaping of the captured data and the motion editing by using the hierarchical solver previously introduced and the de nition of dynamic tasks and constraints. This original method allows adjusting a captured human motion in order to reliably reproduce it on a humanoid while respecting its own dynamics.

Finally, in order to simulate movements resembling to those of humans, we develop an anthropomorphic model with higher number of degrees of freedom than the one of HRP-2. The generic solver is used to simulate motion on this new model. A sequence of tasks is de ned to describe a scenario played by a human. By a simple qualitative analysis of motion, we demonstrate that taking into account the dynamics provides a natural way to generate humanlike movements.

Supervised by N. Mansard, P. Souères, J-Y. Fourquet
Jury: Y. Nakamura (Nakamura Lab, Tokyo Univ.), A. Kheddar (JRL-Japan, AIST-CNRS), V. Padois (ISIR, UPMC), M. Courdesses (LAAS, Univ. Toulouse), P. Soueres (LAAS, CNRS), J-Y Fourquet (ENIT, Univ. Toulouse), N. Mansard (LAAS, CNRS)

Read the manuscript

Footstep planning for humanoid robots: discrete and continuous approaches

By Nicolas Perrin
Defended October the 24th, 2011

Supervised by O. Stasse and F. Lamiraux

Read the manuscript

Optimization and smoothing of task sequences

By Francois Keith
Defended December the 10th, 2010

A general agreed approach on mission and motion planning consists in splitting it into three steps: decomposing the mission into a sequence of tasks (task planning), determining the order of realization and the timing of the tasks (task scheduling) and finally executing the task sequence. This approach maintains the task component in the entire reasoning, using it as a bridge between planning, scheduling and execution.In the sense of task function, a task defines a control law on part of the robot. Hence, for highly redundant systems such as humanoid robots, it is possible to realize several tasks simultaneously using a stack-of-tasks formalism. Though, classical schedulers do not handle the case where the motion is specified not by one, but by a combination of tasks organized into a hierarchy. As a result, the scheduling phase is usually skipped. This thesis aims at reintroducing the scheduling phase, while maintaining the central role of the task.First, the stack-of-tasks formalism is recalled and the continuity of the control law is studied. Particularly, we show that discreet operations (insertion, removal and swap of priority between tasks) create discontinuities in the control. We then present and discuss smoothing methods. Second, we present a task-overlapping based method to optimize not only the scheduling but also the behavior of the tasks of a given sequence, while accounting for the physical constraint of the execution. Finally, we introduce a new perspective in the usage of the task-function approach the task function approach to personalize a task sequence and take into account user preferences.These results are experimented on the humanoid robot platform HRP-2.

keywords: Robotic, Optimization, Scheduling, Task overlapping, Task personalization

Supervised by A. Kheddar, N. Mansard
Jury: C. Chevallereau (IRCCYN, CNRS), M. Beetz (TUM), A. Chrosnier (LIRMM, Univ Montpellier), P-B. Wieber (INRIA Grenoble), A. Kheddar (JRL-Japan, AIST-CNRS), N. Mansard (LAAS, CNRS)

Read the manuscript (in English)