REAL-TIME ARTIFICIAL INTELLIGENCE–BASED VULNERABLE ROAD USERS CLASSIFICATION USING HIGH-FREQUENCY 77 GHZ AUTOMOTIVE RADAR

Vulnerable Road Users (VRUs), such as pedestrians and cyclists, are among the most exposed participants in road traffic environments, and their reliable perception remains a key challenge for Advanced Driver Assistance Systems (ADAS) and autonomous vehicles. Automotive radar operating in the high- frequency 77 GHz band provides robust sensing capabilities with high range and angular resolution, making it well suited for VRU detection under adverse weather and lighting conditions. This paper presents a real-time artificial intelligence–based frame- work for VRU classification using 77 GHz automotive radar measurements. Radar signal processing techniques are applied to extract discriminative kinematic and micro-Doppler features that capture the distinctive motion characteristics of pedestrians and cyclists. These features are used to train a Support Vector Machine (SVM) classifier, selected for its strong generalization capability and low computational complexity, which is critical for real-time automotive applications. The proposed system is evaluated using real-world radar data collected in urban traffic scenarios, demonstrating reliable separation between pedestrian and cyclist classes across varying speeds and motion patterns. The results indicate that the proposed approach provides an effective and computationally efficient solution for real-time VRU classification, supporting its integration into safety-critical ADAS and autonomous driving systems.