Adaptation of Force and Impedance in a Simulated Arm

Abstract

Humans often stabilize objects by co-contraction, also known as the activation of opposing muscles. With more muscle activation there is also more limb stiffness. More muscle activation requires more energy. Several studies suggested a linear learning function for motor learning of movements, based on kinematic error, either in joint or muscle space (for example Kawato, Furukawa, and Suzuki (1987)). However, a V-shaped learning function is preferable, as stated by Franklin et al. (2008). What is the influence of the parameters of the V-shaped learning function on energy and precision? The different values for the parameters influence either energy or precision, so that these two are in a trade-off. In this paper the model proposed by Franklin et al. (2008) is implemented while using generated trajectory data to run tests with varying values for the learning parameters to gain a better insight into this trade-off. It is found that not all parameters correlate with model precision. It is shown that precision increases while at the same time the input energy increases. Therefore we can conclude that there is a trade-off between energy and precision. In general an increase of activation leads to more precision and a decreased efficiency regarding energy.