Research On Dynamic Scheduling And Redistribution Of Emergency Vehicles In Urban Expressway Network

Expressway system, which holds a decisive position in urban traffic operation, has effectively mitigated the traffic pressure in large cities. However, with the continuous increase in the size of vehicle population, frequent traffic congestion and safety in large cities are still prominent problems. Unexpected traffic accidents in an expressway system have significant impacts on the normal operation of urban traffic, and even may pose grave threats people’s life safety. Fast and effective emergency rescue is the key to reduce accident influence, while emergency vehicle scheduling is the core of emergency rescue. The traditional scheduling of emergency vehicles for traffic accidents usually assumes road section weight as statically known, but ignores the effects of changes in road conditions on scheduling work, so in actual scheduling, it is inevitable that an emergency vehicle may suffer a traffic jam and thus cannot ensure the effectiveness of rescue. To minimize the actual response time, it is necessary to set up a multiple factor-based emergency vehicle dynamic scheduling model by making rational use of all available real-time and time-dependent traffic information in accordance with the dynamic characteristics of expressway networks.This dissertation carried out intensive studies on the dynamic scheduling problem of emergency vehicles under the background of multiple traffic accidents in an expressway network. The main research contents and major achievements are shown as follows:(1) Considering the time dependence of expressway networks, this dissertation proposed a time-dependent polygonal-shaped travel speed function to simulate the change of road conditions, and put forward a novel SFLA-based kernel clustering algorithm. The algorithm enhanced the performance of kernel clustering, and realizes an effective fusion of real-time and time-dependent road section speeds. The case study results of Beijing expressway show that the error between the predicted section speed and the actual value was 4km/h.(2) Through analyzing the influence of speed prediction error on route selection, a dynamic shortest path model was set up to optimize the travel time and the route reliability. A shuffled frog-leaping algorithm was put forward to solve the dynamic K shortest paths. The algorithm adopts a random coding scheme to ensure the route connectivity, and then a local optimizing strategy was designed in accordance with the properties of FIFO network shortest paths. This algorithm has an advantage in solving accuracy and speed.(3) With predicted travel time function, accident severity and accident time window as key factors, a route arrangement-based emergency vehicle scheduling model was established. Further, for the influence of unpredictable emergencies on road conditions, taking expressway network nodes as key nodes and a dynamic model for emergency vehicle scheduling is established to dynamically adjust scheduling strategies in line with the real-time data gradually obtained. Considering the weakness of SFLA that it easily runs into local optimization when being used for the solving of complex scheduling problems, an improved SFLA based on the global optimizing capacity of estimation of distribution algorithm is put forward. Compared with SFLA and genetic algorithm, the improved algorithm could help develop a more optimum solution.(4) In consideration of the various regions’competition for resources, the payoff function related to redistribution time and regional potential risk is defined, and a non-cooperative game model for emergency vehicle redistribution is built. After analyzing the hierarchical relation between emergency vehicle scheduling and redistribution, a bi-level programming model is set up. Then a bi-level SFLA with a weight-based SFLA adopted for the frist level, and an SFLA dedicated to non-cooperation game solving adopted for the second level is put forward to solve the model. The case study results of Beijing expressway show that the algorithm has better solving performance than bi-level particle swarm optimization.