2026. A Federated Cost-Aware Deep Reinforcement Learning Framework for Latency-Sensitive and SLA-Compliant Multi-Cloud Auto-Scaling in Distributed Edge Infrastructures. The Science Post. 2, 3 (Sep. 2026).