Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Lightweight Model-Driven 4D Radar Framework for Pervasive Human Detection in Harsh Conditions

Zhenan Liu, Amir Khajepour, George Shaker

TL;DR

This paper presents a fully model-driven 4D radar perception framework designed for real-time execution on embedded edge hardware and suggests that the proposed model-driven approach provides robust, interpretable, and computationally efficient perception for safety-critical applications in harsh industrial and subterranean environments.

Abstract

Pervasive sensing in industrial and underground environments is severely constrained by airborne dust, smoke, confined geometry, and metallic structures, which rapidly degrade optical and LiDAR based perception. Elevation resolved 4D mmWave radar offers strong resilience to such conditions, yet there remains a limited understanding of how to process its sparse and anisotropic point clouds for reliable human detection in enclosed, visibility degraded spaces. This paper presents a fully model-driven 4D radar perception framework designed for real-time execution on embedded edge hardware. The system uses radar as its sole perception modality and integrates domain aware multi threshold filtering, ego motion compensated temporal accumulation, KD tree Euclidean clustering with Doppler aware refinement, and a rule based 3D classifier. The framework is evaluated in a dust filled enclosed trailer and in real underground mining tunnels, and in the tested scenarios the radar based detector maintains stable pedestrian identification as camera and LiDAR modalities fail under severe visibility degradation. These results suggest that the proposed model-driven approach provides robust, interpretable, and computationally efficient perception for safety-critical applications in harsh industrial and subterranean environments.

A Lightweight Model-Driven 4D Radar Framework for Pervasive Human Detection in Harsh Conditions

TL;DR

This paper presents a fully model-driven 4D radar perception framework designed for real-time execution on embedded edge hardware and suggests that the proposed model-driven approach provides robust, interpretable, and computationally efficient perception for safety-critical applications in harsh industrial and subterranean environments.

Abstract

Pervasive sensing in industrial and underground environments is severely constrained by airborne dust, smoke, confined geometry, and metallic structures, which rapidly degrade optical and LiDAR based perception. Elevation resolved 4D mmWave radar offers strong resilience to such conditions, yet there remains a limited understanding of how to process its sparse and anisotropic point clouds for reliable human detection in enclosed, visibility degraded spaces. This paper presents a fully model-driven 4D radar perception framework designed for real-time execution on embedded edge hardware. The system uses radar as its sole perception modality and integrates domain aware multi threshold filtering, ego motion compensated temporal accumulation, KD tree Euclidean clustering with Doppler aware refinement, and a rule based 3D classifier. The framework is evaluated in a dust filled enclosed trailer and in real underground mining tunnels, and in the tested scenarios the radar based detector maintains stable pedestrian identification as camera and LiDAR modalities fail under severe visibility degradation. These results suggest that the proposed model-driven approach provides robust, interpretable, and computationally efficient perception for safety-critical applications in harsh industrial and subterranean environments.
Paper Structure (27 sections, 13 equations, 5 figures, 1 table)

This paper contains 27 sections, 13 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Perception in Adverse Environments
  4. Conventional Automotive Radar
  5. 4D Imaging Radar and Existing Datasets
  6. Classical Geometric Pipelines
  7. Novelty and Positioning of This Work
  8. Methodology
  9. 4D Radar Platform and Multi-Sensor Setup
  10. Domain-Aware Multi-Threshold Filtering
  11. 1) RCS thresholding.
  12. 2) Angular bound filtering.
  13. 3) Doppler plausibility filtering.
  14. Temporal Accumulation with Ego-Motion Compensation
  15. KD-Tree Euclidean Clustering and Doppler-Aware Filtering
  16. ...and 12 more sections

Figures (5)

  • Figure 1: Overview of the model-driven 4D radar pipeline: raw radar measurements through multi-threshold filtering, ego-motion–compensated temporal accumulation, clustering, Doppler-aware refinement, rule-based 3D classification
  • Figure 2: Average point count per sensor under increasing dust concentration. LiDAR returns decline sharply due to attenuation and backscatter, while 4D radar maintains a nearly constant point count with only minor fluctuations, confirming robustness to particulate interference.
  • Figure 3: Qualitative detection results under varying dust conditions. Left: Under clear conditions, IR camera, LiDAR, and radar all provide consistent structure, and both camera-based and radar-based detectors identify pedestrians. Right: At the highest dust concentration, IR and LiDAR fail completely, while the model-driven 4D radar framework continues to detect pedestrians robustly.
  • Figure 4: Pedestrian detections over time from the model-driven radar classifier (red) and a YOLOv8 camera detector (blue). As dust levels increase, camera-based detection collapses, whereas radar detection remains stable.
  • Figure 5: Underground mining evaluation. Camera-based detection fails due to dust and headlamp glare, whereas the model-driven radar framework maintains reliable pedestrian detection.