Evaluating Millimeter-Wave RADAR for Resource-Efficient Drone Perception

Vision is the most widely used perception modality in resource-constrained drones. However, camera-based perception faces limitations in dynamic, cluttered, and dark environments. This paper investigates the application of millimeterwave (mmWave) Radio Detection and Ranging (RADAR) as an Object Detection (OD) and Object Classification (OC) perception modality for small-scale drones due to its resilience to environmental conditions. We propose a lightweight slidingwindow method for static and dynamic OC using sparse RADAR point clouds captured by a cost-effective mmWave RADAR sensor. We compare the performance of RADAR- and visionbased detection performance with dedicated test data for three scenarios. The proposed method achieves robust OD, desired for aerial platforms, effectively handling objects with small RADAR Cross-Section (RCS) of $\mathbf{1 - 1 0} \mathbf{c m}^{\mathbf{2}}$. Comparative analyses against camera-based YOLOv5 show an average recall of 0.8 for three different scenarios. Moreover, our system results in a mean absolute position error between 0.11 m and 0.16 m for RCS values of $\mathbf{1 0}$ and $\mathbf{1} \text{cm}^{2}$ compared to OptiTrack ground truth data. The results show that the proposed solution excels in all scenarios, enabling practical RADAR-based drone perception. Index Terms-perception, object detection, object classifica-