Body intelligence and self-organization of adaptive motion


#1

A well-designed body, with its geometry, materials, dynamics, can considerably simplify sensorimotor control and adaptive interaction with the environment. Understanding scientifically the role of the body in motor control requires the possibility to consider the body as an experimental variable (see http://www.pyoudeyer.com/kaplan-oudeyer-sice09.pdf for a conceptual framework). It has long been very difficult to really explore body variations in a systematic and easy manner. With 3D printing, as Poppy is using, it now becomes possible to consider the body as an experimental variable.

For example, in http://hal.inria.fr/docs/00/86/11/10/PDF/humanoid2013.pdf , we conducted an experiment to study the role of the thigh shape in biped balance and locomotion.

Studying the role of body dynamics is not new at all. Great precursors include researchers like Tad McGeer or Rolf Pfeifer.

An interesting vision is to see the spontaneous body dynamics as a self-organized reservoir of adaptive motions, which can be later on recruited for learning new tasks. A recent and intriguing line of work on this topic is that of the Neurorobotics Lab in Humboldt University (Berlin), which has studied the self-organisation of a collection of simple local controllers embedded on bodies, which can produce adaptive strategies allowing biped legs to stand up very robustly (without models, without learning, only through physical coupling):

http://www.neurorobotik.de/downloads/publications/2011%20Hild%20-%20Attractor-Based%20Behavior%20Control.pdf

http://www.neurorobotik.de/downloads/publications/2011%20Kubisch%20-%20Balance%20Recovery%20of%20a%20Humanoid%20Robot%20Using%20Cognitive%20Sensorimotor%20Loops%20(CSLs).pdf

http://www.neurorobotik.de/downloads/publications/2013%20Hild%20-%20Defying%20Gravity.pdf


Poppy in a musical setup, please share your ideas