Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Benchmarking ML Approaches to UWB-Based Range-Only Posture Recognition for Human Robot-Interaction

Salma Salimi, Sahar Salimpour, Jorge Peña Queralta, Wallace Moreira Bessa, Tomi Westerlund

TL;DR

This work addresses posture estimation for human–robot interaction using ultra-wideband (UWB) range measurements instead of traditional vision or IMU sensors. It introduces a five-node wearable UWB setup and evaluates three supervised classifiers—KNN, SVM, and MLP—for nine-class pose recognition with real-time ROS 2 integration to drive robotic commands. Key findings show SVM and MLP achieve higher accuracy than KNN, while KNN offers faster inference and resilience in certain noise conditions; node-count analysis indicates a minimum of four nodes for reliable performance. The study demonstrates practical applicability through single- and multi-robot experiments (drone and TurtleBot4) and provides a public dataset, highlighting the method’s potential for robust, real-time HRI in challenging environments.

Abstract

Human pose estimation involves detecting and tracking the positions of various body parts using input data from sources such as images, videos, or motion and inertial sensors. This paper presents a novel approach to human pose estimation using machine learning algorithms to predict human posture and translate them into robot motion commands using ultra-wideband (UWB) nodes, as an alternative to motion sensors. The study utilizes five UWB sensors implemented on the human body to enable the classification of still poses and more robust posture recognition. This approach ensures effective posture recognition across a variety of subjects. These range measurements serve as input features for posture prediction models, which are implemented and compared for accuracy. For this purpose, machine learning algorithms including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and deep Multi-Layer Perceptron (MLP) neural network are employed and compared in predicting corresponding postures. We demonstrate the proposed approach for real-time control of different mobile/aerial robots with inference implemented in a ROS 2 node. Experimental results demonstrate the efficacy of the approach, showcasing successful prediction of human posture and corresponding robot movements with high accuracy.

Benchmarking ML Approaches to UWB-Based Range-Only Posture Recognition for Human Robot-Interaction

TL;DR

This work addresses posture estimation for human–robot interaction using ultra-wideband (UWB) range measurements instead of traditional vision or IMU sensors. It introduces a five-node wearable UWB setup and evaluates three supervised classifiers—KNN, SVM, and MLP—for nine-class pose recognition with real-time ROS 2 integration to drive robotic commands. Key findings show SVM and MLP achieve higher accuracy than KNN, while KNN offers faster inference and resilience in certain noise conditions; node-count analysis indicates a minimum of four nodes for reliable performance. The study demonstrates practical applicability through single- and multi-robot experiments (drone and TurtleBot4) and provides a public dataset, highlighting the method’s potential for robust, real-time HRI in challenging environments.

Abstract

Human pose estimation involves detecting and tracking the positions of various body parts using input data from sources such as images, videos, or motion and inertial sensors. This paper presents a novel approach to human pose estimation using machine learning algorithms to predict human posture and translate them into robot motion commands using ultra-wideband (UWB) nodes, as an alternative to motion sensors. The study utilizes five UWB sensors implemented on the human body to enable the classification of still poses and more robust posture recognition. This approach ensures effective posture recognition across a variety of subjects. These range measurements serve as input features for posture prediction models, which are implemented and compared for accuracy. For this purpose, machine learning algorithms including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and deep Multi-Layer Perceptron (MLP) neural network are employed and compared in predicting corresponding postures. We demonstrate the proposed approach for real-time control of different mobile/aerial robots with inference implemented in a ROS 2 node. Experimental results demonstrate the efficacy of the approach, showcasing successful prediction of human posture and corresponding robot movements with high accuracy.
Paper Structure (17 sections, 12 figures)

This paper contains 17 sections, 12 figures.

Table of Contents

  1. Introduction
  2. Methodology
  3. UWB Ranging
  4. Data Acquisition
  5. Machine Learning algorithms
  6. KNN
  7. SVM
  8. MLP
  9. Pose Classification
  10. KNN
  11. SVM
  12. MLP
  13. Experimental Results
  14. Experimental Setup
  15. Obtained Results
  16. ...and 2 more sections

Figures (12)

  • Figure 1: Illustration of the blocks, data flow and use-cases in the presented work. Up to 5 UWB nodes are used to estimate a set of predefined human motions for controlling a single aerial robot or a multi-robot system. We compare the performance of different machine learning algorithms for such task.
  • Figure 2: Showcase of the different classes
  • Figure 3: Each class balanced accuracy
  • Figure 4: Each class $F_{1}$ score
  • Figure 5: KNN confusion matrix obtained with leave-one-out cross-validation (iterating over all subjects).
  • ...and 7 more figures