Control of arm movement using population of neurons

Movements of the human arm in a horizontal plane are very stereotyped in the sense that the corresponding paths are mainly straight lines and the velocity profiles are "bell-shaped like" functions. The dynamics of a two link model of the human arm has been studied with the goal of synthesizing the torques which accomplish the desired transfer. For that purpose a set of parameters which describes the desired transition (initial position, final position, peak velocity, etc.) is chosen randomly according to a certain distribution. The parameters of the desired trajectory as well as the system variables (angles and angular velocities) are encoded using populations of different number of neurons, usually 100-150. The underlying mathematics including integration, differentiation and other algebraic relationships, has been done at the level of neuronal activity. Finally, the driving torques are generated from the corresponding activities using an optimal decoding rule.