|
Understanding Haptics by Building Computational and Physical Models| old_uid | 13739 |
|---|
| title | Understanding Haptics by Building Computational and Physical Models |
|---|
| start_date | 2014/04/01 |
|---|
| schedule | 15h |
|---|
| online | no |
|---|
| details | Séminaire impromptu. Une invitation d'Aymar de Rugy de l'INCIA. |
|---|
| summary | The human hand provides an existence proof that dexterity is possible even as mechatronic systems remain far from such capabilities. It teaches us that hand function in daily life depends on the integration of command signals and contingency plans that anticipate the mechanical dynamics of the limb and its interaction with objects. Conversely, every interaction between the hand and an object provides a wealth of sensory information that is used to update prior hypotheses about the properties and identity of that object and then to select an even better set of command signals and contingency plans to achieve the desired function. Unfortunately, there are no suitable methods to study the neural computation that underlies these capabilities in humans, and dexterous behaviors are limited, difficult to train and difficult to measure in experimental animals. This has led us to develop and test theories of sensorimotor control using computational and mechatronic machines to see if we can replicate biological behaviors. I will review three such models that have generated results that are scientifically interesting and potentially useful:
· A computational model representing the intrinsic mechanical and energetic properties of a two-joint, six-muscle arm plus proprioceptors and spinal interneuronal networks was able to generate the complex coordination required to perform voluntary reaches. Higher motor centers may learn to take advantage of this rather than computing and controlling motor programs directly.
· A compliant fingertip with a contact sensor and spinal-reflex-like control loops greatly improved the usability of a myoelectric prosthetic hand, a form of telerobot.
·A biomimetic tactile sensor combined with a Bayesian algorithm for deciding on exploratory movements outperformed humans in identifying materials based on their surface textures. |
|---|
| responsibles | Deris |
|---|
| |
|