Research And Application Of Urban Optimal Path Planning Based On Spatio-temporal Big Data

In recent years,with the rapid development of the Internet and computer technology,smart mobile devices usually carry position sensors that utilize global positioning technology and generate a large amount of GPS data.This data contains a huge variety of information that can help us better understand moving objects and geographical locations.Among them,timely and accurately recommending the optimal driving path is one of the important problems in traffic data mining.It can guide the direction of traffic flow,facilitate road traffic management,better provide optimal route planning for drivers,and provide reasonable and accurate travel advice for passengers,which plays a crucial role in the design and implementation of urban driving planning.As the traditional problem only considers the network information,while neglecting the needs of users and time-varying external conditions(such as weather,congestion situation),so this article attempt to fuse the taxi GPS trajectory data,the road network structure,static and dynamic information.In the complex urban environment,we present an optimal path planning scheme based on the prediction of driving time.The main work of this paper is as follows:1.Candidate path planning.To establish hierarchical road network path search system based on A* algorithm,the constructed road network topology is extracted.Due to the characteristic of city roads widespread grading,the user’ travel propensity is unknown(such as main route or alley),so the hierarchy is introduced into the path search.The regional road network is arranged as hierarchical,and the direction and distance between the preselected node and the target node in the route are taken as the parameters of the evaluation function,to reduce the number of inflection points in the route.Thus,it is the basis for the subsequent decision to obtain the data of several complete candidate routes.2.Road duration prediction: we construct an end-to-end urban path travel time prediction model,which is the key to innovation and realization of this paper.Firstly,we sample GPS data sequences,based on convolutional neural network during the capture of geographic dependencies.Then consider the future external factors that affect a path time planning: weather,driving condition,regional location,etc.And the metadata for processing is using hot coding with additional input characteristics to explore the influence of external factors.Then they are joined and used as the input of the recursive layer of bidirectional control to obtain accurate time estimation.This paper presents an AB-BiGRU(attention based bidirectional GRU circulation neural network),to the characteristics of the forward and backward transfer vector fusion.Via the full path modeling,the attention mechanism is introduced to enhance the training of special road section,finally be based on running long line optimal candidate set.Then combined with real-time forecast data,user personal orientation(e.g.,least time or shortest distance),and the route restrictions,so as to extract the candidate route in accordance with the reasonable user travel requirement,and present recent rationalization proposal on city planning.Finally,through experiments,the superiority of the route planning model in vehicle navigation is proved by comparing it with other similar models.