Alain Berthoz, Ph.D
College de France
Simplifying Principles for Perception, Action, Locomotion, and Navigation: A Common Problem for Brains and Robots
Wednesday- April 30, 2008
Pacific Forum – 3:00 p.m.
Throughout evolution, simplifying principles have been implemented in biological systems to reduce the complexity of the neural processes necessary to perform rapid and robust sensory-motor tasks or navigation tasks. I shall describe a number of these simplifying principles concerning visual perception, the control of gaze, the generation of hand movements and locomotor trajectories, the mental processes underlying spatial memory for navigation, and manipulation of reference frames. The creation of a fundamental spatial reference frame (the vestibular system) has allowed for the coordination of locomotor activity in a top down manner from the head, used as an inertial stabilization platform.
Natural laws of movement linking the geometry and the kinematics of hand movements are valid also for locomotor trajectories. Recent studies attribute these laws to optimizing principles, such a as minimum jerk, but also propose that they may derive from more general use of non Euclidian geometries. Also, it has been suggested that internal models of the physical properties of moving objects (Newton's law) may allow the brain to anticipate and predict the impact of moving objects during catching tasks.
These simplifying principles may be applied to robotic devices to contribute to the reduction of computational and environmental constraints. The cooperation between roboticians and computer science specialists is essential to Neuroscience for the theoretical contribution brought by these domains which can help to better understand the brain. This talk will be placed in the frame of the search of this fruitful exchange.