Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Online Learning of Wheel Odometry Correction for Mobile Robots with Attention-based Neural Network

Alessandro Navone, Mauro Martini, Simone Angarano, Marcello Chiaberge

TL;DR

An innovative online learning approach for wheel odometry correction is proposed, paving the way for a robust multi-source localization system and shows remarkable results compared to a standard neural network and filter-based odometry Correction algorithms.

Abstract

Modern robotic platforms need a reliable localization system to operate daily beside humans. Simple pose estimation algorithms based on filtered wheel and inertial odometry often fail in the presence of abrupt kinematic changes and wheel slips. Moreover, despite the recent success of visual odometry, service and assistive robotic tasks often present challenging environmental conditions where visual-based solutions fail due to poor lighting or repetitive feature patterns. In this work, we propose an innovative online learning approach for wheel odometry correction, paving the way for a robust multi-source localization system. An efficient attention-based neural network architecture has been studied to combine precise performances with real-time inference. The proposed solution shows remarkable results compared to a standard neural network and filter-based odometry correction algorithms. Nonetheless, the online learning paradigm avoids the time-consuming data collection procedure and can be adopted on a generic robotic platform on-the-fly.

Online Learning of Wheel Odometry Correction for Mobile Robots with Attention-based Neural Network

TL;DR

An innovative online learning approach for wheel odometry correction is proposed, paving the way for a robust multi-source localization system and shows remarkable results compared to a standard neural network and filter-based odometry Correction algorithms.

Abstract

Modern robotic platforms need a reliable localization system to operate daily beside humans. Simple pose estimation algorithms based on filtered wheel and inertial odometry often fail in the presence of abrupt kinematic changes and wheel slips. Moreover, despite the recent success of visual odometry, service and assistive robotic tasks often present challenging environmental conditions where visual-based solutions fail due to poor lighting or repetitive feature patterns. In this work, we propose an innovative online learning approach for wheel odometry correction, paving the way for a robust multi-source localization system. An efficient attention-based neural network architecture has been studied to combine precise performances with real-time inference. The proposed solution shows remarkable results compared to a standard neural network and filter-based odometry correction algorithms. Nonetheless, the online learning paradigm avoids the time-consuming data collection procedure and can be adopted on a generic robotic platform on-the-fly.
Paper Structure (14 sections, 4 equations, 5 figures, 1 table)

This paper contains 14 sections, 4 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Works
  3. Contributions
  4. Methodology
  5. Problem Formulation
  6. Neural Network Architecture
  7. Training Procedure
  8. Tests and Results
  9. Experimental Setting
  10. Evaluation Metrics
  11. Quantitative Results
  12. Latency Evaluation
  13. Conclusions
  14. Acknowledgement

Figures (5)

  • Figure 1: Diagram of the proposed approach. Red blocks and arrows refer to the online training phase, blue ones to the model inference stage, and yellow ones to the odometric input data.
  • Figure 2: Architecture of the proposed model. The batch dimension is omitted for better clarity.
  • Figure 3: Infinite-shaped trajectories estimated by different methods. The data are collected during a total navigation time of about $60 s$.
  • Figure 4: Absolute error of position and orientation of different methods during the test performed on a subset of infinite-shaped trajectories. The considered subset is the same as figure \ref{['fig:fig_xy']}.
  • Figure 5: Histograms of the SE error in position and orientation in section B of the test set.