A McKibben muscle arm learning equilibrium postures

In designing artificial systems for studying motor control in humans and other organisms a key point to consider is the complexity reached by brain and body in their developmental stages. An artificial system whose brain and body complexity is shaped according to developmental stages might allow understanding weather, for example, newborn infants, infants, and adults use different neural mechanisms to cope with the same motor control problems. This article proposes an artificial system which aims at becoming a tool to study this type of problems. The system has a brain and body endowed with a set of minimal bio-mimetic features: (a) neural maps activated by receptive fields; (b) connections plasticity changed by Hebbian rule; (c) robotic arm actuated by a McKibben muscle. The arm autonomously learns to reach specific positions in space under the effect of gravity and for different load conditions. The results suggest that a fast and incremental goal-action mapping formation could constitute the computational mechanism underlying the neural growth and plasticity of an early developed brain at the onset of reaching. The same mechanism also allows a first approximate solution for load compensation avoiding the use of more sophisticated internal models (developed in further brain and body developmental stages). This paper aims to be a preliminary study on the feasibility of this approach.