Investigating the Generalizability of Assistive Robots Models over Various Tasks

TL;DR

This study evaluates the ability of various data-driven models to generalize across different tasks in assistive exoskeletons, highlighting the superior performance of decision tree algorithms and models trained on horizontal plane data.

Contribution

It systematically compares the generalizability of six regression algorithms across multiple tasks, emphasizing the effectiveness of decision trees and horizontal plane models in assistive robotics.

Findings

Decision tree algorithms show the best generalizability.

Models trained on horizontal plane data perform better across tasks.

Task-specific models need less frequent remodeling for new tasks.

Abstract

In the domain of assistive robotics, the significance of effective modeling is well acknowledged. Prior research has primarily focused on enhancing model accuracy or involved the collection of extensive, often impractical amounts of data. While improving individual model accuracy is beneficial, it necessitates constant remodeling for each new task and user interaction. In this paper, we investigate the generalizability of different modeling methods. We focus on constructing the dynamic model of an assistive exoskeleton using six data-driven regression algorithms. Six tasks are considered in our experiments, including horizontal, vertical, diagonal from left leg to the right eye and the opposite, as well as eating and pushing. We constructed thirty-six unique models applying different regression methods to data gathered from each task. Each trained model's performance was evaluated in a cross-validation scenario, utilizing five folds for each dataset. These trained models are then tested on the other tasks that the model is not trained with. Finally the models in our study are assessed in terms of generalizability. Results show the superior generalizability of the task model performed along the horizontal plane, and decision tree based algorithms.