| In the fields of disaster emergency management,one of the most important step is repairing the damaged road network to open up the life passage timely,which makes practical significance in terms of disaster rescue and evacuation of affected people.Most of existing studies focused on the road network,and not considered repair process.In this dissertation,scheduling of road network repair crew is studied based on intelligent decision theory and computer-aided tool,and our main work is as follows:(1)The existing studies of repair crew scheduling are investigated and analyzed,and the main research of this paper are specified.Agent Systems,Markov Decision Process,elements of reinforcement learning problems,and Q-learning algorithm are introduced,which can provide solution for repair crew scheduling.(2)A model and solution for non-continuous damaged roads repairing is proposed.A mathematic model is constructed to describe the network with non-continuous damaged roads,where the damaged roads are represented by damaged nodes and all non-demand nodes are removed.The Markov Decision Process is adopted to model the repair process,and action space,state space,and reward function is designed for the Agent.Q-learning is used to solve the repair crew scheduling in the network with non-continuous damaged roads.Comparison experiments show that our algorithm is more robust,can obtain more effective and reasonable schedules in various network conditions with high transport and repair efficiency.(3)A model and solution for continuous damaged roads repairing is proposed.For the situation with continuous damaged roads,partial non-demand nodes are considered,and a decision-making model for the repair crew scheduling is constructed,which is proved to be a Markov Decision Process.A Q-learning based scheduling algorithm is proposed for continuous damaged roads repairing.Experiments show that even in the most serious and complicated situations of the road network,the algorithm can solve the schedule problem effectively. |
