Rendering realistic weight sensations in virtual reality with personalized electrical muscle stimulation

RWTH Publications (RWTH Aachen)

Abstract

Deep immersion in virtual reality (VR) and augmented reality (AR) is limited by the absence of convincing haptic feedback, particularly the sensation of weight. While Electrical Muscle Stimulation (EMS) offers a promising solution, a systematic framework for its effective and personalized application is lacking. This thesis addresses this gap through a multi-study investigation to develop and validate methods for rendering realistic weight sensations using personalized EMS within VR. A foundational study (N=75) demonstrated that a dual-muscle EMS strategy is significantly superior to single-muscle approaches for perception and realism, establishing the benefit of synergistic muscle actuation. A second study (N=35) then systematically examined moderating factors, revealing that ergonomic parameters, such as a 90° elbow angle, are critical for realism. This study also identified a perceptual optimum for weight augmentation (+2 kg) and confirmed that individual user characteristics (e.g., age, BMI) significantly influence perceived realism, indicating that "one-size-fits-all" approaches are inadequate. Finally, a third study (N=36) indicated that for greater realism, EMS-based haptic feedback should be congruently synchronized with visual cues, as incongruent stimuli were found to create a significant sense of mismatch. To address the need for personalization, machine learning models were developed using data from the first study. These models achieved F1-scores of up to 77% in predicting perceived weight differences based on key features including applied EMS amplitude, user thresholds, and BMI. In synthesis, this thesis presents a multi-tiered framework for the optimization and personalization of EMS-based weight rendering. By identifying the effectiveness of synergistic muscle actuation, quantifying ergonomic and individual factors, defining the principles of visuohaptic congruence, and developing data-driven models for personalization, this work establishes design guidelines for immersive VR/AR systems. While these findings are bounded by factors such as the specific muscle groups and quasi-static holding tasks investigated, the developed methodology provides a validated basis for adapting haptic feedback to individual user characteristics.

Authors 1

1. Apostolos Vrontos corresponding Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen