Figure 2
From: An analytical method reduces noise bias in motor adaptation analysis
Overview of the movement generation framework during a task of tossing a ball to hit a target. In the Planning phase, the thrower generates a motor command \((u)\) that, in a noise-free environment, will result in a specific ball landing point \(\left({x}_{u}\right).\) In other words, a control action \(u\) is formed using an inverse model of the user's estimate of system dynamics \((\widehat{f}(u))\), and \({x}_{u}\) is obtained by propagating this action \(u\) through the actual dynamics \(f\): [\({x}_{u}{=f({\widehat{f}}^{-1}(x}_{i}))\)]. The difference between \({x}_{u}\) and the intended target \(({x}_{i})\) represents misestimation of system parameters that are continually updated through the learning process. In the Movement phase, the throw is completed with \({x}_{u}\) being affected by control noise \(({\eta }_{q})\) to produce the actual measurable position \(({x}_{m})\). In the Sensing phase, the actual movement endpoint \(({x}_{m})\) is corrupted by feedback noise \(({\eta }_{r}\)), resulting in the endpoint sensed by the thrower \(({x}_{s})\). In the Perceiving phase, a posterior estimate \(({x}_{p})\) of the landing point is arrived at by combining information from the intended endpoint \(\left({x}_{i}\right)\), the sensed endpoint \(({x}_{s})\), and the level of internal model noise \((\xi )\)1,11.
