Research On Path Planning Method Of Urban Sanitation Vehicle In Time-varying Traffic Environment

In recent years,the process of building smart city in China is gradually accelerating.As an important part of smart city,smart environmental sanitation is also attracting more and more attention.Relying on the Internet plus sanitation management platform,it is expected to realize the informationization and refinement of sanitation management.At present,there are more information services and less intelligent decision support in the existing environmental sanitation related application systems.In this context,this thesis studies the decision-making theory of environmental sanitation vehicle path planning in time-varying environment.Sanitation vehicles have been playing an important role in urban streets,including street cleaning,dust removal and cooling,and garbage collection.There are time-varying traffic flow,traffic accidents,weather changes and other uncertain factors in urban streets,which affect the operation efficiency of sanitation vehicles.Once sanitation vehicles work for a long time,it will bring more burden to the traffic in the city.Therefore,it is very important to plan the route reasonably and reduce the operation time of sanitation vehicles while considering the time-varying traffic flow or uncertain factors.Based on the sanitation vehicle street cleaning operation scenario,this thesis studies two kinds of variants which are a basic operational research problem:(1)This thesis studies the time-dependent rural postman problem(TDRPP)of sanitation vehicles in time-varying traffic flow scenarios.Considering the requirements of sanitation vehicles serving specific streets,the shortest time goal and the time dependence of traffic flow,TDRPP is studied to find the shortest off-line path planning scheme of sanitation vehicles.In this thesis,a mathematical model of TDRPP is established based on the time-space network model.This model adopts the idea of discrete-time network and arc path alternation.Compared with the existing modeling method based on finite piecewise constant function,it can more accurately describe the time-varying characteristics of TDRPP.In addition,an optimal solution property of TDRPP is discussed in time-dependent networks satisfying the first in first out(FIFO)property.Based on this optimal property,a special genetic algorithm(GA)is designed to solve large-scale TDRPP effectively,and the solution time of the proposed GA is further reduced by a bridging path caching strategy.Finally,a large number of time-varying networks are simulated to verify the effectiveness of the proposed method.(2)This thesis studies the stochastic time-dependent rural postman problem(STDRPP)of sanitation vehicles under time-varying traffic flow and uncertain factors.On the basis of TDRPP,the on-line path planning scheme of sanitation vehicle is further considered after the traffic accident,weather mutation and other uncertain factors cause the travel time to be extended.STDRPP is modeled as a markov decision process(MDP),and an adaptive model training method based on deep q-network(DQN)is designed.The experimental results show that the trained decision-making model can adaptively identify uncertain factors and carry out subsequent path planning online.Compared with offline path planning method and online greedy method,the model can effectively reduce the operation time of sanitation vehicles in time-varying network.In summary,this paper studies the path planning theory of sanitation vehicles under time-varying traffic environment,which has certain theoretical guiding significance.