Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Dispelling Four Challenges in Inertial Motion Tracking with One Recurrent Inertial Graph-based Estimator (RING)

Simon Bachhuber, Ive Weygers, Thomas Seel

TL;DR

This research extends the Recurrent Inertial Graph-based Estimator to generalize across a large range of sampling rates, and demonstrates that RING can overcome four real-world challenges: inhomogeneous magnetic fields, sensor-to-segment misalignment, sparse sensor setups, and nonrigid sensor attachment.

Abstract

In this paper, we extend the Recurrent Inertial Graph-based Estimator (RING), a novel neural-network-based solution for Inertial Motion Tracking (IMT), to generalize across a large range of sampling rates, and we demonstrate that it can overcome four real-world challenges: inhomogeneous magnetic fields, sensor-to-segment misalignment, sparse sensor setups, and nonrigid sensor attachment. RING can estimate the rotational state of a three-segment kinematic chain with double hinge joints from inertial data, and achieves an experimental mean-absolute-(tracking)-error of 8.10 +/- 1.19 degrees if all four challenges are present simultaneously. The network is trained on simulated data yet evaluated on experimental data, highlighting its remarkable ability to zero-shot generalize from simulation to experiment. We conduct an ablation study to analyze the impact of each of the four challenges on RING's performance, we showcase its robustness to varying sampling rates, and we demonstrate that RING is capable of real-time operation. This research not only advances IMT technology by making it more accessible and versatile but also enhances its potential for new application domains including non-expert use of sparse IMT with nonrigid sensor attachments in unconstrained environments.

Dispelling Four Challenges in Inertial Motion Tracking with One Recurrent Inertial Graph-based Estimator (RING)

TL;DR

This research extends the Recurrent Inertial Graph-based Estimator to generalize across a large range of sampling rates, and demonstrates that RING can overcome four real-world challenges: inhomogeneous magnetic fields, sensor-to-segment misalignment, sparse sensor setups, and nonrigid sensor attachment.

Abstract

In this paper, we extend the Recurrent Inertial Graph-based Estimator (RING), a novel neural-network-based solution for Inertial Motion Tracking (IMT), to generalize across a large range of sampling rates, and we demonstrate that it can overcome four real-world challenges: inhomogeneous magnetic fields, sensor-to-segment misalignment, sparse sensor setups, and nonrigid sensor attachment. RING can estimate the rotational state of a three-segment kinematic chain with double hinge joints from inertial data, and achieves an experimental mean-absolute-(tracking)-error of 8.10 +/- 1.19 degrees if all four challenges are present simultaneously. The network is trained on simulated data yet evaluated on experimental data, highlighting its remarkable ability to zero-shot generalize from simulation to experiment. We conduct an ablation study to analyze the impact of each of the four challenges on RING's performance, we showcase its robustness to varying sampling rates, and we demonstrate that RING is capable of real-time operation. This research not only advances IMT technology by making it more accessible and versatile but also enhances its potential for new application domains including non-expert use of sparse IMT with nonrigid sensor attachments in unconstrained environments.
Paper Structure (11 sections, 4 equations, 4 figures, 2 tables)

This paper contains 11 sections, 4 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Methods
  4. Simulated Training Data at Various Sampling Rates
  5. Neural Network Architecture: RING with Sampling Rate Input
  6. Results and Discussion
  7. Experimental Setup and Data Acquisition
  8. Evaluation Metrics and Baselines
  9. Experimental Validation of RING
  10. Real-time Applicability of RING
  11. Conclusion

Figures (4)

  • Figure 1: A three-segment KC with two IMUs (blue boxes). The graph representation of the KC is given by the parent array $\boldsymbol{\lambda} = (0, 1, 2)^\intercal$. The neural network-based multiple-IMU sensor fusion algorithm RING receives the graph representation and IMU data as input and estimates the rotational state of the KC, overcoming all four key challenges of IMT simultaneously.
  • Figure 2: Experimental 3D-printed KC used to validate the RING algorithm. To validate that RING overcomes the IMT challenge of non-rigid sensor placement, each segment of the KC has an IMU attached using foam padding. Additionally, a second IMU is rigidly attached to assess the impact of the foam padding on the accuracy of orientation estimates. Figure from bachuberRINGicml24.
  • Figure 3: Experimental example sequence that demonstrates RING's prediction performance from sparse, magnetometer-free, nonrigidly attached IMUs and without joint axes direction, and comparing to ground truth orientations for the first 15 of one trial and for a double hinge joint KC with joint axes directions in $x-$ and $y-$direction.
  • Figure 4: Experimental magnetometer-free motion tracking accuracy (in degrees) of RING across various sampling rates. RING achieves a nearly constant estimation accuracy across sampling rates ranging from 50.0 to 200. Uncertainties are one standard deviation.