 |
|
|
|
 |
|
|
|
|
|
The Lab. has been created in March 2009.
We conduct researches in the field of Robotics. More particularly we are focusing on finding dynamics features that characterize humans and can be understood by robots. We develop formalisms and methodologies to understand human and human motions from their dynamics, as well as to measure humanoids dynamics. This will help in developing more friendly and personalized robotics systems, and will enhance human-robot interaction.
The range of applications includes humanoid robotics, human/robot interction, medical diagnostic support, rehabilitation monitoring, sport science, entertainment...
On-going research
Around the above problematics we are now actively doing research on the following areas (partners):
- Optimal human body modelling for simulation (Toyota Central Laboratory, Nagoya, Japan)
- Identification of the human body dynamics for applications in Biomechanics (Laboratoire Complexite, Innovation et Activites Motrices et Sportives, Paris University, Orsay, France)
- Computation of the contact forces of legged systems, without forceplate, application to rehabilitation and gait evaluation (National Rehabilitation Center, Tokorozawa, Japan)
- Affect recognition from motion and biometrics (Laboratoire de Physiologie et de Perception de l'Action, College de France, Paris, France)
- Systematic identification of dynamics for legged systems
- Adaptative control of robot for optimal and safe interaction with human (Tagawa Lab., TUAT, Tokyo, Japan)
- Marker-less motion capture develpoment (Nakamura Lab., TUAT, Tokyo, Japan)
- Dynamics modeling of human, and muscle identification (INRIA, Montpellier, France & Nakamura Lab., Tokyo University, Japan)
- Identification theory (IRCCyN, Nantes France)
- Evaluation of fatigue from motion (Adaptative Systems Laboratory, Waterloo University, Canada
- Foot modelling
State of the art
Identification of legged system dynamics
A classic identification approach requires the measurement of the all the joint torques, all the joint angles, the generalized coordinates, the contact forces. It also requires the modeling of the joint visco-elasticity and friction. The measurement of the joint torque is often difficult, with important errors, and the modelling of joint properties is still an open problem. The general approach we proposed is based on the decomposition of the inverse dynamics of a mobile system into the base-link equations describing the 6 DOF of the base-link in the 3D space, and the equations of the kinematics. By making use of the base link dynamics only we have proven that the dynamics is fully identifiable and does not require the joint torques, neither the modelling of the joint properties. Identification of the pure dynamic parameters can be achieved with a minimal of measurements.This method applies particularly for legged systems and to humanoid robots. We have tested the method with humanoid robot of different sizes.
| An other application is the identification of the human dynamics. Using a multi-body rigid model of the human we can measure with accuracy the mass parameters of each segment. A real time application that allows the visualization of the results as been developped. Such method allow to monitor the segment parameters during rehabilitation or physical training. A marathon runner has undergo the measurements during 5 months of training, and parameters have been monitored successfully. |
 |
Identification of human joint dynamics
Pressing needs in developing tools to support medical diagnostic, and more particularly neuro-muscular disease diagnostic and their follow-up have conducted us in developing a method to identify the passive visco-elasticity of limb joints.
This method allows to classify patients in accordance to the visual assessment and to assign a figure to the viscoelastic parameters. The passive properties are modeled with a biomedical model of the joints passive properties, and each of the parameters for each joint is identified simultaneously using the captured motion data during the diagnosis.
|
 |
 |
Identification of muscle dynamics
To understand human motion generation and control, muscle modeling is crucial. So far the Hill-Stroeve model has been the most widely used, however literature parameters characteristic of the model are not fit to describe subject specificities. Based on musculo-skeletal computations, we have identified the Hill-Stroeve parameters of the elbow flexion and extension muscles for individuals and showed that the muscle force estimation is dramatically changed using the appropriate parameters. |
From now on
Identification of legged system dynamics
A lighter measuring environment is required to roll-out the identification methodology. One of our research direction is to replace the motion capture system by smaller systems. Marker-less systems, stereo-vision, gyro-sensors etc...
|
People recognition and understanding of emotions from motions
Motions dynamics is typical to an individual. Our preliminary results have shown that from motions it is possible to recognize a person and develop algorithms to achieve that goal.
Motions convey a lot of information, not only purely mechanical. Our goal is to extract as many information as possible from motions and particularly the one related to emotions. We have developped an algorithm that can recognize between 4 types of emotions: neutral, happiness, sadness and anger. These researches are conducted in collaboration with Prof. Berthoz from College de France, Paris, France and Dr. Kadone from University of Tsukuba.
Muscle dynamics
Research on muscle dynamics and identification of muscle models are also an active field of research in collaboration with Dr. Hayashibe amd PRof. Fraisse in Montpellier, France.
Complex system modeling, identification and control
In collaboration with Prof. Tagawa Laboratory, we conduct research in novel approaches to control complex mechanical systems such as humanoid robots and parallel systems (earthquake simulators, driving simulators...) based on dynamics modeling.
last update 01.07.2011 by g*
copyright Gentiane Venture 2009
|
|
