# Modeling human visuomotor adaptation with a disturbance observer framework

## TL;DR

This paper proposes a new model for how the brain adapts hand movements when visual errors are introduced, using a control theory framework.

## Contribution

The paper introduces a novel discrete-time state space model of visuomotor adaptation based on the internal model principle and a disturbance observer.

## Key findings

- The DO Model incorporates a disturbance observer and a feedforward system to explain visuomotor adaptation.
- The model addresses key requirements like modular architecture and abstraction capacity.
- The paper suggests that distinct instructions in visuomotor experiments can reveal different motor learning components.

## Abstract

A fundamental problem of visuomotor adaptation research is to understand how the brain is capable to asymptotically remove a predictable exogenous disturbance from a visual error signal using limited sensor information by re-calibration of hand movement. From a control theory perspective, the most striking aspect of this problem is that it falls squarely in the realm of the internal model principle of control theory. Despite this fact, the relationship between the internal model principle and models of visuomotor adaptation is currently not well developed.

This paper aims to close this gap by proposing an abstract discrete-time state space model of visuomotor adaptation based on the internal model principle. The proposed DO Model, a metonym for its most important component, a disturbance observer, addresses key modeling requirements: modular architecture, physically relevant signals, parameters tied to atomic behaviors, and capacity for abstraction. The two main computational modules are a disturbance observer, a recently developed class of internal models, and a feedforward system that learns from the disturbance observer to improve feedforward motor commands.

A central research challenge in the field of motor learning is to disentangle the components of motor learning [1–8]. The main hypothesis explored in this paper is that visuomotor rotation experiments in which participants are given distinct instructions to either move the hand or the cursor to the target offer a platform to explore two components of motor learning arising from different computations and possibly different brain regions. The two computations correspond to a disturbance observer whose role is to detect and then eliminate environmental perturbations and a feedforward system whose role is to learn from the disturbance observer to improve feedforward motor commands.