Meeting Abstract

93.1  Monday, Jan. 6 13:30  Optimization for models of legged locomotion: parameter estimation, gait synthesis, and experiment design BURDEN, SA*; SASTRY, SS; FULL, RJ; Univ. of California, Berkeley

Optimization theory provides a powerful framework for the study of organismal biomechanics that enables: estimation of uncertain model parameters or exogenous inputs using empirical data; synthesis of gaits and dynamic maneuvers that extremize a given performance criteria; and design of experiments to maximally distinguish predictions of competing models or sets of hypotheses. Such techniques are well-studied for continuous models that are specified by a single ordinary differential equation, but less well-developed for the piecewise-defined or discontinuous models that naturally describe terrestrial locomotion. Computationally-tractable algorithms for parameter estimation near a periodic gait undergoing a prespecified sequence of isolated foot touchdown events exist, but still impose stringent assumptions on the sequence of discrete events or the region of the state or parameter space that may be explored. We propose a scalable algorithm for optimization of piecewise-defined models for terrestrial locomotion that generalizes previous efforts by defining and numerically computing the first-order variation of model trajectories. We apply nonlinear programming to solve optimization problems while imposing minimal restrictions on model structure or event sequences. In particular, our method can optimize footfall sequence and timing for aperiodic gaits. More broadly, this foundational technique for parameter estimation, gait synthesis, and experiment design provides a link between model-based investigations and data-driven experimental study of legged locomotion.