Modeling Rhythmic Biological Behaviors Using Limited Data


Ankaralı M. M. , Nicole D., Tytell E., Cowan N.

Sıam Conference On The Life Sciences, Massachusetts, United States Of America, 11 - 15 July 2016

  • Publication Type: Conference Paper / Summary Text
  • City: Massachusetts
  • Country: United States Of America

Abstract

A staggering diversity of biological and engineered systems exhibit rhythmic behaviors, and their dynamics have been analyzed for hundreds of years. Yet, data-driven modeling and analysis of rhythmic behaviors remains in its infancy and such tools are essential for understanding systems for which “first-principles” models are not feasible. Identifying the dynamics of rhythmic systems from input–output data is critical to many applications in robotics and biology, and yet remains a challenge. Here, we describe a new formulation for identifying rhythmic dynamical systems by using harmonic transfer functions. This formulation side-steps the well-known problem of estimating the limit cycle itself, enables separate identification of input and measurement delays, and applies to both hybrid and continuous dynamical systems. An important feature of our work is the selection of effective stimuli when large numbers of oscillatory cycles cannot be easily obtained. We present preliminary results on the application of these techniques to the identification of muscle dynamics in experiments involving ex vivo lamprey muscle.