The human movement system is extraordinarily complex, but at an abstract level, its function can be characterized as manipulating its Newtonian dynamics to achieve useful goals.
Generating the dynamics to make this happen takes advantage of three main system organizations:
1) the architecture of its skeleton, that constraints its degrees of freedom,
2) the modulation of its muscles' stiffnesses, that allows the control of movement variability, and
3) its neural feedback systems, which generate torques through an elaborate system of spinal reflexes. Everyday movements manipulate all these factors to combine postural stability with dexterous manipulation.
Although the goals behind some tasks can be described succinctly, to achieve them the system must control approximately six hundred muscles acting about three hundred joints. This overall complexity has limited the study of movement to relatively simple situations such as walking, cycling and reaching, that can be analyzed with approximate tractable models that have much lower degrees of freedom. Such studies show marked regularities across subjects that can be interpreted in in terms of relating the dynamics of the movement to its goals.
However, it is still an open question as to whether regularities extend to arbitrary large-scale movements. Do humans use a common control strategy or do movements vary significantly across different individuals? To address this question, we analyzed posture variability when individuals traced predetermined paths on a large scale that required them to make whole body movements with steps. Subjects traced with their index finger at a controlled speed in a virtual three-dimensional environment.
The principal result is 11:52 13.11.2018that although they were not instructed as to the details of their movement, subjects exhibited remarkably similar posture sequences throughout the movements, as well as common modulations of variability in different degrees of freedom. The extraordinary similarities in posture control raise the possibility that an endogenous Bayesian principal based on frequency of use may extend to all movements.
If this turns out to be the case it may lead to a common movement control principle.
About the Speaker:
Dana Ballard graduated from MIT with a bachelors degree in aeronautics and astronautics. He then attended the University of Michigan for his masters in information and control engineering in 1970. He received his Ph.D. from the University of California, Irvine in information engineering in 1974. He has done research in artificial intelligence and human cognition and perception with a focus on the human visual system. In 1982, with Christopher M. Brown he authored a pioneering textbook in the field of computer vision, titled Computer Vision.
He is also known as a proponent of active vision techniques for computer vision systems as well as approaches to understanding human vision. His textbook titled „An Introduction to Natural Computation“ (1997) combines material on varied subjects relevant to computing in the brain, such as neural networks, reinforcement learning, and genetic learning. His most recent book, „Brain Computation as Hierarchical Abstraction,“ describes a multilevel approach to understanding neural computation.