Better Science Through Predictive Modeling: Numerical Tools for Understanding Neuromechanical Interactions

Abstract

The animal kingdom is filled with amazing examples of coordinated locomotor and balance behavior. The intricate interaction of the neuromechanics of the combined skeletal, muscular, and neural systems that underlie these behaviors only adds to their impressiveness. To wit, the neuromechanics must deal with fantastically nonlinear dynamics, delayed and noisy sensory input, and multiple stability regimes in unpredictable environments. Because of these underlying complex interactions, an integrative systems approach is required to understand the performance of the locomotor and balance behavior that emerges. In this chapter, we propose the use of predictive modeling to facilitate the investigation of neuromechanics using our software platform, Neuromechanic. With this technique the dynamics of constituent neuromechanical systems are modeled and the resulting emergent behaviors studied; holistic behaviors are an output rather than an input for simulation. We describe three ways in which software can aid in a predictive approach to neuromechanical modeling: first, use of tools that emphasize control and optimization for predictive modeling; second, visualization and organization to aid in careful parameterization necessary to account for the variation found in biological specimens; third, building confidence in modeling results through the use of sensitivity analysis. We offer examples of these techniques using Neuromechanic, which is designed to simplify the prototyping of neural control strategies, formulate optimization criteria, visualize key parameters that effect model performance, and succinctly perform sensitivity analysis.