YoloRL: simplifying dynamic scheduling through efficient action selection based on multi-agent reinforcement learning Konferenz-Paper