International Conference on Computer and Knowledge Engineering

Home / 15th International Conference on Computer and Knowledge Engineering

Deep Learning-based Processing of Autonomous Vehicle Radar Data to Achieve High Resolution

Authors :

Nima Abdolrahimi Shahamat¹ Vahideh Moghtadaiee² Esfandiar Mehrshahi³

1- Faculty of Electrical Engineering, Shahid Beheshti University, Tehran, Iran 2- Cyberspace research institute, Shahid Beheshti University, Tehran, Iran 3- Faculty of Computer Science and Engineering, Shahid Beheshti University

Keywords :

Automotive Radar،Deep Learning،Radar-only Perception،Object Classification،Autonomous Driving

Abstract :

This research investigates deep learning methods for classifying moving objects in autonomous driving using automotive radar data. Radar sensors are compact, efficient, and robust to adverse weather, making them attractive for standalone perception. We design and evaluate three radar-only models Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and a hybrid RNN-LSTM trained and tested exclusively on radar features. These radar-only models achieve up to 88.27\% accuracy (RNN-LSTM) and 84.53\% (LSTM), confirming that radar alone can effectively support object detection and classification. In addition, we introduce a CNN+Image variant in which training is performed jointly on radar and image data but inference relies on radar only, reflecting real-world deployment where cameras may be unavailable or unreliable. This cross-modal training strategy yields improved radar-only inference accuracy of 89.23\%, showing that image information during training can enhance radar feature learning even when only radar is used at test time. Overall, this study demonstrates the potential of radar-based deep learning as a practical and cost-effective solution for autonomous vehicles.