Research On Model And Algorithms Of The Traffic Evacuation Planning For City Emergency

The population in the city is dense, when fatal disaster come, how to evacuate the thick public in a large area in the shortest possible time is an important safely measure to avoid or reduce casualties. The major of the evacuation route planning is to calculate the optimum route planning strategy, follow which the whole time needed for all the evacuees to arrive at the destinations is as short as possible.In the research field of the evacuation for regional area, some research aimed to evaluate the evacuation time without artificial control as accurate as possible, others tried to optimize the evacuation time under the artificial control, and calculated some control strategy. For example, somebody calculated the maximum-flow route, and then suggests all the vehicles try to evacuate through this route; some researcher proposed the idea that all the evacuees try to evacuate through the nearest exit, and think that would be efficient; someone planned all the static evacuation routes on the standard that the intersections are least, etc. The defect is:follow all these static methods, the route resources are difficult to utilized fully, the traffic jams would likely to be caused, so is difficult to achieve the optimized evacuation time.Our algorithm can provide specific evacuation routes and arrive time at every node for every evacuee. We consider both time and space as resources, adopted the ideas of pre-allocating the resources, to avoid the congestion, and achieve the optimization of the overall evacuation time. The strategy of giving up optimal solution, but calculating the sub-optimal solution, gives a possibility of practical application for the large-scale network.At first, the evacuation problem is abstracted as a network model that composed of nodes and edges. The node has the attribute of capacity, and the edge with the capacity and travel time attributes. The problem is to transport all units in source nodes to destination nodes, and the objective is the shortest evacuate time.In order to get the network model abstracted from actual road information, the methods to model the node and edge are researched detailed in the model part of the paper, in which, the attribute of the edge is mainly got from the analysis and conversion of the road traffic capacity, the attribute of the node is mainly comes from the model and analysis of the signalized intersection traffic capacity. To get the source node and the evacuee amount on which, the method to model the initial population distribution is studied.After established the network model, the solution of the evacuation route planning problem is researched. In the classic field of the network flow, the original problem should first be transformed to static network flow problem like minimum cost flow problem using the time-expand-graph, then it can be calculated by some polynomial time algorithms. But the time-expand-graph needs huge spending to extend the original network to the whole evacuation time times, and deterministic algorithms need too much cost when calculating the huge networks. For these reasons, a kind of heuristic algorithm is studied in this paper, which don't ensure to get the optimal solution, but suboptimal solution, with advantage that the resource consumption and calculate time can be reduced significantly, so the solving of the practical large network becomes possible. Besides to carry out the algorithm and improved the pre-processing part, which can simplify the programming and reduce the task in each iteration, and this change doesn't affect the result, then this paper improved the original algorithm from two different point. First, the iterative threshold value is devised to terminate iterations earlier for every departure time, in order to eliminate some long path form evacuation route plan. Second, we let the evacuees longer from the destinations start first, to reduce the arrival time of last one.Besides, the result of the algorithm is demonstrated on the GIS flat root, and the framework and flow for the application to the real time traffic control system is discussed.