Integrating federated learning and YOLOv11 for object detection in autonomous vehicles

Autonomous vehicles (AVs) require high-performance systems for environmental perception, especially for accurate object detection. Systems based on deep neural networks perform well, but often face challenges due to the high computational cost and the need for centralised data collection. This paper proposes the integration of federated learning with the YOLOv11 model to create more efficient and scalable ATS solutions. The FLYolo11 model is presented, which optimises object detection in computationally constrained environments, improving accuracy and performance without the need for centralised training. Experimental results show that the model significantly improves the detection performance at average computational cost compared to other approaches.