latency-minimization

Proposes a two-timescale multi-layer deep reinforcement learning framework with latent action space for joint service placement, computational delegation, and power control in hierarchical edge-cloud computing, achieving up to 20.8% latency reduction and 13% resource utilization improvement.