Travel Time Prediciotn Based On Massive Taxi Trajectory Data

In recent years,the number of vehicle has increased dramatically in China while the construction of road infrastructure has relatively slow growth,which intensifies the contradiction between the traffic supply and demand,and causes frequent traffic congestion in high-urbanized area.In such circumstances,travel time of traveler is complex and changeable,and travel cost becomes higher.Accurate travel time prediction is undoubtedly of importance to both traffic managers and travelers,and in the meantime,predicting travel time is never a straightforward task considering the dynamics of traffic situations.But fortunately,mature intelligent transportation system(ITS)and emergence of massive trajectory data give an unrivalled opportunity to reach accurate travel time prediction.In this context,this paper attempts to explore travel time prediction in high-urbanized area based on massive vehicle trajectory data.Though vehicle trajectory data can supply complete traffic information in spatial and temporal aspect,its "big" volume brings difficulties to the maintenance and the retrieval of traffic information.In addition,the accuracy of travel time prediction is not only depend on the performance of prediction model,but also limited by the complexity of travel time series.Then,before predicting travel time,the complexity of travel time series in the aspect of prediction need to be explored and the predictability of travel time need to be evaluated.Based on above,travel time prediction should include several aspects,i.e.trajectory data modeling,trajectory data indexing,travel time predictability measuring,and travel time predicting,etc.For achieving them,this paper defines vehicle trip data model to storage vehicle trajectory data,designs vehicle-based trip indexing scheme to access to trip information efficiently,measures the predictability of historical travel time series to examine and evaluates the effect of input data of travel time prediction to predicted results,and,as the final target,develop travel time predicting model involved various traffic factors to predict travel time.Therefore,this paper focuses on trajectory data modeling,trajectory data indexing,travel time predictability measuring,and travel time predicting to achieve accurate travel time prediction.In terms of trajectory data modeling,this paper uses traffic flow direction-based road network model to model complicated urban road networks contained express road and general urban road,proposes "vehicle-based trip" data model to organize vehicle trajectory data,and develops topology-based map matching algorithm to match vehicle trajectory."vehicle-based trip" denotes a travel experience of traveler described by vehicle trajectory data.To extract trip information from vehicle trajectory data,this paper defines vehicle-based trip as the unit of storage of trajectory data to integrate trajectory data and express trip information,and creates "vehicle-based trip" set from trajectory data using trajectory extracting,map matching,trip splitting and other steps.For matching trajectory data to road network,this paper develops a novel topology-based map matching algorithm.It filters an alternative road segments set of a trajectory series by the shortest distance principle,and uses the topologies of road segments to calculate the feasible path set from the alternative road segment set,then select the matched path from the feasible path set based on inertia principle,expressway-first principle,and the shortest path principle,finally,corrects the location of every GPS sample points of trajectory data and establishes the connection between vehicle trajectory and road network model.Empirical research introduces the complete process of modeling trajectory data,demonstrates the availability of "vehicle trip" data model and the proposed topology-based map matching algorithm,and discusses the performance of modeling trajectory data.To trajectory data index,the very large data volumes of massive trajectory data make the traditional index mode in database management systems not completely efficient to maintain and use all-purpose indexing structures.This paper introduces a data indexing scheme to specific massive spatial-temporal data.Its principle is to develop an indexing structure with respect to the characteristics of the data and the application purposes.In particular,the approach is implemented and applied to vehicle trajectory data in the city of Shanghai including more than 100 million locational records per day collected from about 13,000 taxis.In order to search for patterns and trends across the traffic data generated at any time between for instance any entrance and exit in the Shanghai express road system,this paper developed a case-based indexing structure,named "TripCube".The TripCube is computationally compared to all-purpose indexing structures,and it appears that TripCube improves performances to a great extent.Next,this paper discusses the predictability of travel time.Based on the analysis of the complexity of travel time series,this paper defines travel time predictability to express the probability of correct travel time prediction and proposes an entropy-based method to measure the upper bound of travel time predictability.Multiscale entropy is employed to quantify the complexity of travel time series,and the relationships between entropy and the upper bound of travel time predictability are presented.Empirical studies are made with vehicle trajectory data in an express road section.The effectiveness of time scales,tolerance,and series length to entropy and travel time predictability are analysis,and some valuable suggestions about the accuracy of travel time predictability are discussed.Finally,the comparisons between travel time predictability and actual prediction results from two prediction models,ARIMA and BPNN,are conducted.Experimental results demonstrate the validity and reliability of the proposed travel time predictability.Based on above research works,this paper develops a trip-oriented travel time prediction method.In highly-urbanized areas,trip-oriented travel time prediction(TOTTP)are valuable to travelers rather than traffic managers as the former usually expect to know the travel time of a trip which may cross over multiple road sections.There are two obstacles to the development of TOTTP,including traffic complexity and traffic data coverage.With large scale historical vehicle trajectory data and meteorology data,this research develops a BPNN-based approach through integrating multiple factors affecting trip travel time into a BPNN model to predict trip-oriented travel time for OD pairs in urban network.Results of experiments demonstrate that it helps discover the dominate trends of travel time changes daily and weekly,and the impact of weather conditions is non-trivial.At the end,the research findings of this paper are summeried,major contributions and limitations of our research works are listed,and their future prospects are presented.