Edge Radar Material Classification Under Geometry Shifts

TL;DR

This paper introduces a lightweight mmWave radar-based material classification system for edge devices, highlighting its high accuracy under ideal conditions and significant performance drops under geometry shifts, with analysis and potential solutions.

Contribution

It presents a novel radar material classification pipeline optimized for edge devices, and provides an in-depth analysis of robustness issues caused by geometry shifts.

Findings

Achieves 94.2% macro-F1 under nominal geometry

Performance drops to 68.5% macro-F1 under geometry shifts

Identifies systematic intensity scaling and RCS effects as failure modes

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.