Tweet
The LAAS lab at the department of humanoid robot (http://projects.laas.fr/gepetto/index.php), University of Toulouse (France) invites applications for fully funded Ph.D. position commencing in September or october 2014.
Context
This PhD project presents a unique interdisciplinary approach based on biomechanics and robotics in order to study human movement organization involving full body motion, to model it and simulate its variability.
State of the Art
The human body has a large number of joints and muscles which allows him to perform a wide range of tasks. Although human movement is studied for a long time, his organization is still poorly understood. How the central nervous system does to succeed in selecting and in coordinating the activity of a set of muscles to realize a given movement? The difficulty of this question formulated in the middle of the last century [Lashley 33, Bernstein 67], is induced from the fact that each movement is the result of a particular command among an infinite number. However, the central nervous system is able to find extremely fast and robust solutions to the problem of muscles and kinematic redundancies and produces stable and accurate movements. To date, no computational model allowed developing an algorithm of motion planning with performance comparable to that of human subjects in terms of speed, precision, robustness and adaptability. The reason is certainly that the good criteria of motor control have not yet been identified.
Objectives of the research project
During this project, we propose to analyze human movement taking into account its variability by choosing motor tasks which involve full body motion using possible dynamic constraints with the environment. These movements will be multi-tasks and will obey to a set of constraints more or less explicit. After definition of reference tasks entailing full body motion, a first part of the work, falling within the biomechanics, will involve observation and analysis of movement from human subjects. For studying these movements the Phd student will develop a dynamic model of the human musculoskeletal system that will allow us to calculate joint torques and to estimate muscles forces.
At the same time, algorithms for motion planning developed by robotics will be used to simulate upon dynamic models of anthropomorphic systems, the same reference tasks as those performed by the human subjects. For this, we will use advanced software using "stack of tasks" to generate full-body motion movements coming from a hierarchy of goals and constraints described in the form of algebraic linear equalities and inequalities and powerful numerical solvers. This software developed for humanoid robotics, can create movement sequences with large variability (using selection of function tasks, defining constraints, configuring optimization criteria used by the solver, etc). These settings are now selected and set by the robotics to ensure the proper execution of the task based on selected performance criteria. With this study, we hope to simulate for any given sequence the variability of the movement that we observed in human.
Results coming from the observation of human movement and the motion planning of anthropometric models will be then analyzed and compared. By performing coming and going between experimental and mock-up approaches, we will determine optimal parameters in order to generate real movements and to explore its variability. On this basis, each approach will feed the information provided by the other.
Requirements and proposal
The candidate should have a Master Degree in Mathematics or Mechanical engineering with a solid Biomechanics or Robotics backgrounds.
For an informal discussion on the proposed project, please contact Bruno Watier (bruno.watier@univ-tlse3.fr) or Philippe Soueres (soueres@laas.fr). French description of the project could be find here: http://projects.laas.fr/gepetto/uplo...Candidat-1.pdf
To apply, please send a one page motivation letter, and a detailed CV.
Context
This PhD project presents a unique interdisciplinary approach based on biomechanics and robotics in order to study human movement organization involving full body motion, to model it and simulate its variability.
State of the Art
The human body has a large number of joints and muscles which allows him to perform a wide range of tasks. Although human movement is studied for a long time, his organization is still poorly understood. How the central nervous system does to succeed in selecting and in coordinating the activity of a set of muscles to realize a given movement? The difficulty of this question formulated in the middle of the last century [Lashley 33, Bernstein 67], is induced from the fact that each movement is the result of a particular command among an infinite number. However, the central nervous system is able to find extremely fast and robust solutions to the problem of muscles and kinematic redundancies and produces stable and accurate movements. To date, no computational model allowed developing an algorithm of motion planning with performance comparable to that of human subjects in terms of speed, precision, robustness and adaptability. The reason is certainly that the good criteria of motor control have not yet been identified.
Objectives of the research project
During this project, we propose to analyze human movement taking into account its variability by choosing motor tasks which involve full body motion using possible dynamic constraints with the environment. These movements will be multi-tasks and will obey to a set of constraints more or less explicit. After definition of reference tasks entailing full body motion, a first part of the work, falling within the biomechanics, will involve observation and analysis of movement from human subjects. For studying these movements the Phd student will develop a dynamic model of the human musculoskeletal system that will allow us to calculate joint torques and to estimate muscles forces.
At the same time, algorithms for motion planning developed by robotics will be used to simulate upon dynamic models of anthropomorphic systems, the same reference tasks as those performed by the human subjects. For this, we will use advanced software using "stack of tasks" to generate full-body motion movements coming from a hierarchy of goals and constraints described in the form of algebraic linear equalities and inequalities and powerful numerical solvers. This software developed for humanoid robotics, can create movement sequences with large variability (using selection of function tasks, defining constraints, configuring optimization criteria used by the solver, etc). These settings are now selected and set by the robotics to ensure the proper execution of the task based on selected performance criteria. With this study, we hope to simulate for any given sequence the variability of the movement that we observed in human.
Results coming from the observation of human movement and the motion planning of anthropometric models will be then analyzed and compared. By performing coming and going between experimental and mock-up approaches, we will determine optimal parameters in order to generate real movements and to explore its variability. On this basis, each approach will feed the information provided by the other.
Requirements and proposal
The candidate should have a Master Degree in Mathematics or Mechanical engineering with a solid Biomechanics or Robotics backgrounds.
For an informal discussion on the proposed project, please contact Bruno Watier (bruno.watier@univ-tlse3.fr) or Philippe Soueres (soueres@laas.fr). French description of the project could be find here: http://projects.laas.fr/gepetto/uplo...Candidat-1.pdf
To apply, please send a one page motivation letter, and a detailed CV.