Proper performance index is selected, and each scheduling decision is produced and improved in simulation with NDP.

• 选择合理的性能指标，采用神经元动态规划产生每一步的调度，并在仿真中优化策略。