IMU-Based Activity Recognition and Payload Estimation for Augmentative Exoskeleton

Augmentative exoskeleton is a wearable device which is used to enhance human physical capabilities by providing additional support and strength. Such exoskeleton uses a multi-tier control system to automatically adjust its support based on the user’s activity and the payload’s weight. The high-level control system establishes the foundation for adaptive control responses by analyzing the wearer’s movements and payload conditions. The mid-level controller maps these activities into precise control commands in form of torque and angular velocity at the joints of the exoskeleton. At the low level, a feedback controller executes these commands. In this study, a high-level control system is developed to utilize the data collected from the Inertial Measurement Unit (IMU) to recognize crucial exoskeleton movements such as walking, lifting, and lowering, along with estimating the weights of the carried payload. Most of the existing studies primarily focuses on activity recognition, leaving payload estimation relatively unexplored, which is equally crucial for developing such controller. This study proposes a hybrid deep learning framework that combines a Convolutional Neural Network (CNN) with a Bidirectional Gated Recurrent Unit (BiGRU), enhanced with an attention mechanism. The proposed model performs not only activity recognition but also the estimation of the carried payload’s weight while improving upon traditional methods with an F1-score of 95.14% for the activity recognition and 95.86% for the payload estimation on a publicly available open-source dataset.

• Publication year

  2025

• Venue

  International Conference on Robotics and Artificial Intelligence

• Publication date

  2025-12-19

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1109/ICRAI68431.2025.11396723
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Analysis of Machine Learning-based Human Activity Prediction Model for Assistive Exoskeleton

  2024cited by this paper
• Human Activity Recognition and Payload Classification for Low-Back Exoskeletons Using Deep Residual Network

  2023influential reference
• IMU-based human activity recognition and payload classification for low-back exoskeletons

  2023influential reference
• Wearable Sensor-Based Human Activity Recognition with Hybrid Deep Learning Model

  2022cited by this paper
• A Comparative Analysis of Hybrid Deep Learning Models for Human Activity Recognition

  2020cited by this paper
• Industrial exoskeletons: Need for intervention effectiveness research.

  2019cited by this paper
• Mobile sensor data anonymization

  2018influential reference
• Sensor Data Preprocessing, Feature Engineering and Equipment Remaining Lifetime Forecasting for Predictive Maintenance

  2018cited by this paper
• Control strategies for active lower extremity prosthetics and orthotics: a review

  2015cited by this paper
• Adam: A Method for Stochastic Optimization

  2014cited by this paper
• Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

  2014cited by this paper
• A Public Domain Dataset for Human Activity Recognition using Smartphones

  2013influential reference
• Personalization and user verification in wearable systems using biometric walking patterns

  2012influential reference
• Leading Edge of Cybernics: Robot Suit HAL

  2006cited by this paper
• Long Short-Term Memory

  1997cited by this paper

• No citing papers are available for this paper.