Recreating life-like motion in robotic limbs

A study in which the motion of living limbs was characterized and formulated for the control of mechanical limbs is discussed. Trajectories of human elbow joints performing a blind reaching motion were digitized and smoothed using an A/D converter on a personal computer. Waveforms for position over time were recorded and velocity and acceleration waveforms derived from them. Trajectories were normalized with respect to time and amplitude and a polynomial curve and its first two derivatives fitted to them. Finally, the human trajectories were compared to that of a robotic arm controlled by a biologically based algorithm using artificial nerve cells.<>