Computational Models of Human Movement
- Cet article est une traduction de :
- Modèles computationnels du mouvement humain
Dinesh Pai, Professor at the University of Vancouver (Canada), invited by the Assembly of the Professors on the proposition of Prof Alain Berthoz.
He gave in May and June 2009 four lectures entitled:
Computational Models of Human Movement
1. Creating virtual objects that look, feel, and sound real
2. Automated capture of human movement and object behaviour
3. What robots teach us about human movement
4. Modeling the neurobiology of human movement
1The theme of this lecture series is how we can understand human movement by exploiting new developments in computing. In the course of the four lectures I will show how computer science has embarked on an exciting new journey to understand physical and physiological reality. Specifically, new computer simulation methods with unprecedented levels of realism, combined with new imaging and measurement methods that allow us to create virtual environments in which the complexity of human movement could be understood.
2Humans move and experience the real world by exploiting all their available senses, including vision, touch, and hearing. This is because real objects respond to human interaction in a multisensory manner, by resisting touch, moving, changing shape, and making sounds. Therefore, virtual environments should also provide such correlated multisensory information that is both realistic and responsive to human interaction. This first lecture will introduce recent developments in my laboratory and elsewhere for creating multisensory virtual worlds with computer graphics, haptics, and auditory displays. I will describe the dramatic changes in computer hardware, and the efficient new algorithms for interactive simulation of contact between humans the physical environment. In addition to applications in computer games and movies, these developments allow realistic study of how humans interact with their environment.
3To build computer simulations that are realistic and useful it is essential measure real behavior. Traditionally these types of measurements were tedious and time-consuming. In this second lecture, I will describe modern imaging and measurement systems that make measurement-based modeling much more tractable. Medical imaging techniques, particularly MRI, make it possible to acquire accurate anatomical and behavioral models of human subjects in unprecedented detail. High speed motion capture systems enable accurate measurement of complex human movement such as facial expressions. Finally, new robotic systems that can probe physical objects can automatically acquire behavioral models. I will give examples from work in my laboratory to illustrate how this is done. The examples include the UBC Active Measurement Facility (ACME), a telerobotic facility for ``reality-based’‘ modeling physical properties of objects, including its shape, appearance, deformation response, sound, and surface roughness.
4Much of our current understanding of human movement is descriptive. To have a deeper understanding, it is important to appreciate the physical constraints on any organism, whether human or robot, that must successfully interact with the physical world. As Horace Barlow observed: “A wing would be a most mystifying structure if one did not know that birds flew”. Indeed, classical qualitative knowledge about feathers and flapping, known to Icarus, did not provide the keenest insight; rather, it was the modern quantitative knowledge of aerodynamic lift, turbulence, stability, and control. The third lecture will describe how building robots can give us similar insights into the requirements for successful movement, and provide a rigorous test of our understanding of human movement. Using robotic hands and walking robots as examples, I will describe how the successes and failures of robotics are teaching us what is really important about human movement.
5The final lecture will describe computational models of the complex interplay between neurons, muscles, bones, sensors, and other tissues that produce human movement. Combining recent developments in multisensory computer simulation, novel measurement techniques, and the insights from robotics described in the previous lectures, I will show how we can build detailed models of the human sensorimotor system and its interaction with the physical world. I will describe the beginnings of this quantitative approach, focusing on two areas of my current research: the control of eye movement and manipulation with hands. These computational models promise not only deeper insights into how the brain controls movement but also new applications to human health.
Pour citer cet article
The Letter of the Collège de France, Collège de France, Paris, juin 2009, p. 30. ISSN 1958-1408
Dinesh Pai, « Computational Models of Human Movement », La lettre du Collège de France [En ligne], 4 | 2008-2009, mis en ligne le 15 novembre 2010, consulté le 24 avril 2017. URL : http://lettre-cdf.revues.org/762Haut de page
© Collège de FranceHaut de page