Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Edge Radar Material Classification Under Geometry Shifts

Jannik Hohmann, Dong Wang, Andreas Nüchter

Abstract

Material awareness can improve robotic navigation and interaction, particularly in conditions where cameras and LiDAR degrade. We present a lightweight mmWave radar material classification pipeline designed for ultra-low-power edge devices (TI IWRL6432), using compact range-bin intensity descriptors and a Multilayer Perceptron (MLP) for real-time inference. While the classifier reaches a macro-F1 of 94.2\% under the nominal training geometry, we observe a pronounced performance drop under realistic geometry shifts, including sensor height changes and small tilt angles. These perturbations induce systematic intensity scaling and angle-dependent radar cross section (RCS) effects, pushing features out of distribution and reducing macro-F1 to around 68.5\%. We analyze these failure modes and outline practical directions for improving robustness with normalization, geometry augmentation, and motion-aware features.

Edge Radar Material Classification Under Geometry Shifts

Abstract

Material awareness can improve robotic navigation and interaction, particularly in conditions where cameras and LiDAR degrade. We present a lightweight mmWave radar material classification pipeline designed for ultra-low-power edge devices (TI IWRL6432), using compact range-bin intensity descriptors and a Multilayer Perceptron (MLP) for real-time inference. While the classifier reaches a macro-F1 of 94.2\% under the nominal training geometry, we observe a pronounced performance drop under realistic geometry shifts, including sensor height changes and small tilt angles. These perturbations induce systematic intensity scaling and angle-dependent radar cross section (RCS) effects, pushing features out of distribution and reducing macro-F1 to around 68.5\%. We analyze these failure modes and outline practical directions for improving robustness with normalization, geometry augmentation, and motion-aware features.
Paper Structure (23 sections, 7 figures)

This paper contains 23 sections, 7 figures.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. METHODOLOGY
  4. Data Acquisition and Hardware Configuration
  5. Feature Extraction
  6. Range Bin Selection
  7. Intensity Descriptor
  8. Model Architecture and Edge Deployment
  9. EXPERIMENTAL EVALUATION
  10. Baseline Performance under Nominal Geometry
  11. Classification Accuracy and Confusion Analysis
  12. Confidence Distribution
  13. Performance under Geometry Shifts
  14. Impact of Sensor Height Displacement
  15. Impact of Inclination Angle and RCS
  16. ...and 8 more sections

Figures (7)

  • Figure 1: Default experimental setup used for baseline data acquisition. The radar is mounted in a nadir-facing configuration at $H=45$ cm and $\theta=90^\circ$ relative to the material surface.
  • Figure 2: Confusion matrix of the MLP classifier under nominal conditions ($H=45$ cm, $\theta=90^\circ$). Diagonal entries indicate per-class recall.
  • Figure 3: Confidence (maximum Softmax probability) distribution under nominal geometry.
  • Figure 4: Confusion matrix under height variations (35 cm and 55 cm). Off-diagonal density increases due to distance-dependent power changes.
  • Figure 5: Confidence distribution under geometry shifts (tilt). Increased low-confidence predictions indicate reduced certainty under OOD geometry.
  • ...and 2 more figures