The Effects of Different Motor Teaching Strategies on Learning a Complex Motor Task

Tjasa Kunavar, Marko Jamsek, Edwin Johnatan Avila Mireles, Elmar Rueckert, Luka Peternel, Jan Babic

TL;DR

This paper investigates how different robotic teaching strategies (non-assisted, assisted, and segmentation) affect learning a complex, visuomotor-perturbed motor task. While all strategies led to learning, the segmentation approach showed an initial advantage in performance on the main task, but no significant long-term benefits in generalization compared to simpler methods.

What Problem Does This Paper Address?

Learning new motor skills often involves a "teaching-learning" interaction, where an expert guides a novice. While haptic devices have advanced our understanding of how individuals acquire skills, the specific strategies for delivering instructional stimuli and information haven't been extensively explored. This paper aims to fill this gap by examining how different robotic teaching methods impact learning a challenging, novel motor task that minimizes reliance on prior knowledge. The task involves a visuomotor perturbation, meaning the visual feedback of the movement doesn't directly correspond to the physical arm movement, requiring significant adaptation and learning.

Methods

The researchers designed a novel tracking task with a visuomotor perturbation. Participants used a haptic robot to control a virtual pointer on a screen. The core of the task involved a continuously rotating mapping between the arm's movement and the pointer's movement, creating a complex visuomotor challenge.

Three experimental groups were formed, each employing a different teaching strategy:

  • Non-Assisted (NAD): The robot acted solely as an interface, providing no guidance. This served as the control group.
  • Assisted (ATD): The robot provided intermittent, mild haptic guidance (a spring-like force) to help the pointer follow the target.
  • Segmentation (SEG): Subjects first learned two simpler sub-tasks (Rotation and Translation) before performing the main task, also with intermittent haptic assistance. This strategy aimed to break down the complex task into more manageable components.

Performance was evaluated based on accuracy (how close the pointer was to the target) and max error (the maximum Euclidean distance between the pointer and target centers).

Datasets

No external datasets were used. The study generated its own experimental data from the participants performing the designed motor task. The "dataset" consists of the recorded performance metrics (scores and errors) of the 27 participants across multiple trials and sessions.

Key Insights and Ideas

  • Novelty of the Task: The visuomotor perturbation task is designed to be challenging and minimize the influence of prior motor knowledge, making it a good testbed for studying learning strategies.
  • Initial SEG Advantage: The segmentation group showed a performance advantage when first introduced to the main task, suggesting that learning sub-components can prime performance for the complex task.
  • No Long-Term Advantage for Segmentation: Despite the initial boost, the segmentation strategy did not lead to superior performance or better generalization compared to the assisted or non-assisted groups in the long run.
  • Generalization is Tricky: The study highlights that generalization of learned skills to new, albeit related, scenarios (like varying rotation angles) is complex and may require more than just breaking down a task.
  • Importance of Task Design: The paper emphasizes that the design of the motor task itself is crucial for studying learning and that the metrics used (score and max error) are effective for tracking learning progress.

Key Results

  • All three groups (NAD, ATD, SEG) successfully learned the complex motor task, showing improvement in scores and reduction in errors over the three sessions.
  • In the first session, the SEG group had significantly higher average scores and lower max errors when first exposed to the Main Task compared to NAD and ATD groups.
  • However, by the end of the training phase, there were no significant differences in average scores or max errors between the groups for the Main Task.
  • During the generalization phase, while all groups improved, there were no consistent significant differences in performance or error reduction between the groups across sessions. The NAD group showed slightly better performance in the second generalization session.
  • Performance in the generalization phase was inversely proportional to the rotation angle, meaning higher rotation angles led to lower scores.

Limitations and Future Work

  • Limited Haptic Assistance: The assisted group only received a simple spring-like haptic force, lacking other sensory feedback (visual or auditory) that might enhance learning.
  • Hardware Limitations: The force applied for assistance might have been too low to significantly impact learning for the ATD group.
  • Segmentation Complexity: The segmentation approach might have been unintuitive for participants, or the specific sequence of sub-tasks might not have been optimal.
  • Gaming Experience: The study acknowledges that participants' prior experience with video games could influence their ability to adapt to virtual environments.
  • Future Work: The authors suggest exploring the impact of gaming experience, individualizing teaching strategies based on participant responses, and investigating how to improve generalization and long-term retention. They also propose using VR headsets for greater immersion and developing more robust mathematical models for sensorimotor teaching.

Summary updated

Abstract

During the learning of a new sensorimotor task, individuals are usually provided with instructional stimuli and relevant information about the target task. The inclusion of haptic devices in the study of this kind of learning has greatly helped in the understanding of how an individual can improve or acquire new skills. However, the way in which the information and stimuli are delivered has not been extensively explored. We have designed a challenging task with nonintuitive visuomotor perturbation that allows us to apply and compare different motor strategies to study the teaching process and to avoid the interference of previous knowledge present in the naïve subjects. Three subject groups participated in our experiment, where the learning by repetition without assistance, learning by repetition with assistance, and task Segmentation Learning techniques were performed with a haptic robot. Our results show that all the groups were able to successfully complete the task and that the subjects’ performance during training and evaluation was not affected by modifying the teaching strategy. Nevertheless, our results indicate that the presented task design is useful for the study of sensorimotor teaching and that the presented metrics are suitable for exploring the evolution of the accuracy and precision during learning.
BibTeX
@inproceedings{kunavar2024effects,
  author = {Tjasa Kunavar and Marko Jamsek and Edwin Johnatan Avila Mireles and Elmar Rueckert and Luka Peternel and Jan Babic},
  title = {{The Effects of Different Motor Teaching Strategies on Learning a Complex Motor Task}},
  booktitle = {Sensors},
  year = {2024},
  url = {https://openreview.net/forum?id=cgCLzKlyoc},
}

Loading related content...

Figure