Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A machine learning framework for acoustic reflector mapping

Usama Saqib, Letizia Marchegiani, Jesper Rindom Jensen

TL;DR

This work presents a ML-augmented acoustic reflector mapping framework that combines a Sequential Non-linear Least Squares TOA estimator, MPDR DOA estimation, and an RBF-SVM classifier to distinguish true echoes from noise in noisy, reverberant indoor environments. The approach targets robust mapping under $-10$ dB SNR and varying $T_{60}$, aiming to enable reliable echo-based localization on resource-constrained robots. Experimental results in simulated rooms show $TOA$ accuracy around $80\%$ at $-10$ dB and competitive computation times, with the SVM classifier reducing spurious wall detections and improving spatial map quality. The method offers a practical pathway to integrate acoustic sensing into multi-modal navigation systems where cameras or lidars are challenged.

Abstract

Sonar-based indoor mapping systems have been widely employed in robotics for several decades. While such systems are still the mainstream in underwater and pipe inspection settings, the vulnerability to noise reduced, over time, their general widespread usage in favour of other modalities(\textit{e.g.}, cameras, lidars), whose technologies were encountering, instead, extraordinary advancements. Nevertheless, mapping physical environments using acoustic signals and echolocation can bring significant benefits to robot navigation in adverse scenarios, thanks to their complementary characteristics compared to other sensors. Cameras and lidars, indeed, struggle in harsh weather conditions, when dealing with lack of illumination, or with non-reflective walls. Yet, for acoustic sensors to be able to generate accurate maps, noise has to be properly and effectively handled. Traditional signal processing techniques are not always a solution in those cases. In this paper, we propose a framework where machine learning is exploited to aid more traditional signal processing methods to cope with background noise, by removing outliers and artefacts from the generated maps using acoustic sensors. Our goal is to demonstrate that the performance of traditional echolocation mapping techniques can be greatly enhanced, even in particularly noisy conditions, facilitating the employment of acoustic sensors in state-of-the-art multi-modal robot navigation systems. Our simulated evaluation demonstrates that the system can reliably operate at an SNR of $-10$dB. Moreover, we also show that the proposed method is capable of operating in different reverberate environments. In this paper, we also use the proposed method to map the outline of a simulated room using a robotic platform.

A machine learning framework for acoustic reflector mapping

TL;DR

This work presents a ML-augmented acoustic reflector mapping framework that combines a Sequential Non-linear Least Squares TOA estimator, MPDR DOA estimation, and an RBF-SVM classifier to distinguish true echoes from noise in noisy, reverberant indoor environments. The approach targets robust mapping under dB SNR and varying , aiming to enable reliable echo-based localization on resource-constrained robots. Experimental results in simulated rooms show accuracy around at dB and competitive computation times, with the SVM classifier reducing spurious wall detections and improving spatial map quality. The method offers a practical pathway to integrate acoustic sensing into multi-modal navigation systems where cameras or lidars are challenged.

Abstract

Sonar-based indoor mapping systems have been widely employed in robotics for several decades. While such systems are still the mainstream in underwater and pipe inspection settings, the vulnerability to noise reduced, over time, their general widespread usage in favour of other modalities(\textit{e.g.}, cameras, lidars), whose technologies were encountering, instead, extraordinary advancements. Nevertheless, mapping physical environments using acoustic signals and echolocation can bring significant benefits to robot navigation in adverse scenarios, thanks to their complementary characteristics compared to other sensors. Cameras and lidars, indeed, struggle in harsh weather conditions, when dealing with lack of illumination, or with non-reflective walls. Yet, for acoustic sensors to be able to generate accurate maps, noise has to be properly and effectively handled. Traditional signal processing techniques are not always a solution in those cases. In this paper, we propose a framework where machine learning is exploited to aid more traditional signal processing methods to cope with background noise, by removing outliers and artefacts from the generated maps using acoustic sensors. Our goal is to demonstrate that the performance of traditional echolocation mapping techniques can be greatly enhanced, even in particularly noisy conditions, facilitating the employment of acoustic sensors in state-of-the-art multi-modal robot navigation systems. Our simulated evaluation demonstrates that the system can reliably operate at an SNR of dB. Moreover, we also show that the proposed method is capable of operating in different reverberate environments. In this paper, we also use the proposed method to map the outline of a simulated room using a robotic platform.
Paper Structure (14 sections, 16 equations, 4 figures)

This paper contains 14 sections, 16 equations, 4 figures.

Table of Contents

  1. INTRODUCTION
  2. Related Works
  3. Problem Formulation
  4. Sequential Non-linear Least Squares (S-NLS)
  5. DOA Estimation
  6. Acoustic Echoes Classifier
  7. Implementation Details
  8. The Beamformer
  9. The SVM Classifier
  10. Experimental Evaluation
  11. Experimental Setup
  12. Results
  13. Application Example
  14. Discussion and Conclusion

Figures (4)

  • Figure 1: a) TOA and b) DOA evaluation under different SNRs
  • Figure 2: a) TOA and b) DOA evaluation under different room reverberation
  • Figure 3: Spatial Map of an indoor environment with dimension $8 \times 6 \times 5$ m a) without classifier b) with SVM classifier
  • Figure 4: Spatial Map of an indoor environment with dimension $13 \times 11 \times 5$ m a) without classifier b) with SVM classifier