Learning postures through sensorimotor training: a human simulation case study. Epigenetic Robotics

In this paper, we present a case study showing why the modeling effort of computa-tional neuro-sciences must call upon interac-tions with an environment when developmen-tal phenomena are investigated. More specif-ically, in the context of a model of the spatial organization of a repertoire of postures, we show that calling upon a realistic simulation of the human kinematics (i) results in raising the question of the nature of information en-coded in this repertoire; and (ii) reveals that some previous assumptions about the func-tional organization of this repertoire were not necessary. Finally, we point out that con-versely, robotics can benefit from such stud-ies by using the suggested principles to design more adaptive control architectures. 1