基于偏好函数与深度信念网络的自动化码头AGV实时调度算法

为提高自动化集装箱码头岸桥作业效率、降低自动导引车(Automated Guided Vehicle,AGV)空载行驶距离,以偏好函数和深度信念网络(Deep Belief Network,DBN)为核心设计AGV实时调度算法,完成AGV任务分配指派,优化AGV作业任务序列。利用偏好函数为处于待分配状态的AGV筛选最优集装箱,产生训练样本并更新训练集,通过DBN实时更新集装箱的偏好函数,重复该过程直至集装箱作业完毕。算例分析表明:同两种现行调度规则对比,AGV空载距离和岸桥作业时间显著下降;与人工神经网络(Artificial Neural Network,ANN)相比,DBN可有效提高算法精度与求解效率;算法针对环境动态变化实时分配集装箱,为自动化码头提高效率和降低能耗提供依据。

Real-Time Scheduling Algorithm for AGV in Automated Container Terminals Based on Preference Function and Deep Belief Net

The AGV(Automated Guided Vehicle)scheduling algorithm is devised based on the preference function and the deep belief net to increase the efficiency of quay cranes and decrease the unloaded distance of AGV in automated container terminals.The algorithm performs AGV job assignment and optimizes the AGV job sequence.The preference function is used to sort out the most favorable container for the AGV to be allocated,to generate training samples and to update the training set.The preference function is updated by DBN(Deep Belief Network)throughout the operation.Numerical experiments indicate that compared with two existing scheduling rules,the proposed algorithm can increase the efficiency of quay cranes and decrease the unloaded distance of AGV.The DBN can effectively improve algorithm accuracy and efficiency compared with ANN(Artificial Neural Network).The algorithm dispatches containers in real time according to the dynamic changes of the situation,which provides a basis for the improvement of efficiency and the reduction of energy consumption in automated container terminals.