أرسل هذا في رسالة قصيرة: Federated learning for digital twin applications: a privacy-preserving and low-latency approach