Abstract
This paper presents a numerical modeling technique for dynamically modeling a human hand. We use a strand-based method of modeling the muscles. Our technique represents a compromise between capturing the full dynamics of the tissue mechanics and the need for computationally efficient representations for control design and multiple simulations appropriate for statistical planning tools of the hand. We show how to derive a strand-based model in a variational integrator context. Variational integrators are particularly well-suited to resolving closed-kinematic chains, making them appropriate for hand modeling.We demonstrate the technique first with a detailed exposition of modeling an index finger, and then extend the model to a full hand with 19 rigid bodies and 23 muscle strands. We end with a discussion of future work, including the need for impact handling, surface friction representations, and system identification.