Cusumano Joseph P, Mahoney Joseph M, Dingwell Jonathan B
Department of Engineering Science & Mechanics, Penn State University, University Park, PA, 16802, USA,
Adv Exp Med Biol. 2014;826:125-45. doi: 10.1007/978-1-4939-1338-1_9.
Using the concept of task manifolds, a number of data analysis methods have been used to explain how redundancy influences the structure of variability observed during repeated motor performance. Here we describe investigations that integrate the task manifold perspective with the analysis of inter-trial task dynamics. Goal equivalent manifolds (GEMs), together with optimal control ideas, are used to formulate simple models that serve as experimentally testable hypotheses on how inter-trial fluctuations are generated and regulated. In an experimental context, these phenomenological models allow us to show how error-correcting control is spatiotemporally organized around a given GEM. To illustrate our approach, we apply it to study the variability observed in a virtual shuffleboard task. The geometric stability properties of the inter-trial dynamics near the GEM are extracted from fluctuation time series data. We find that subjects exhibit strong control of fluctuations in an eigendirection transverse to the GEM, whereas they only weakly control fluctuations in an eigendirection nearly, but not exactly, tangent to it. We demonstrate that our dynamical analysis is robust under coordinate transformations, and discuss how our results support a generalized interpretation of the minimum intervention principle that suggests the involvement of competing costs in addition to goal-level error minimization.
利用任务流形的概念，人们使用了多种数据分析方法来解释冗余如何影响重复运动表现过程中观察到的变异性结构。在此，我们描述了将任务流形视角与试验间任务动力学分析相结合的研究。目标等效流形（GEMs）与最优控制思想一起，用于构建简单模型，这些模型可作为关于试验间波动如何产生和调节的可实验验证假设。在实验环境中，这些现象学模型使我们能够展示纠错控制是如何在给定的GEM周围进行时空组织的。为了说明我们的方法，我们将其应用于研究虚拟推圆盘游戏任务中观察到的变异性。从波动时间序列数据中提取GEM附近试验间动力学的几何稳定性特性。我们发现，受试者对与GEM横向的特征方向上的波动表现出强有力的控制，而对几乎但不完全与GEM相切的特征方向上的波动控制较弱。我们证明了我们的动力学分析在坐标变换下是稳健的，并讨论了我们的结果如何支持对最小干预原则的广义解释，该解释表明除了目标水平的误差最小化之外，还涉及相互竞争的成本。
