Target Classification through ISAR for Autonomous Vehicles based on Federated Learning

This study explores the use of Federated Learning (FL) in classifying ISAR images for autonomous driving. Automotive radar systems, operating at millimeter-wave frequencies, offer critical safety features. ISAR images are powerful for target recognition but pose challenges in real-world scenarios. FL, a decentralized training approach, is employed for data privacy while maintaining competitive accuracy. Our findings reveal that FL achieves commendable performance compared to centralized models, ensuring data confidentiality by keeping the information on local devices and centrally sharing only the model weights. In conclusion, this research demonstrates FL's potential in improving ISAR-based target classification for autonomous driving, making it suitable for privacy-sensitive applications.