| Short-term forecasting of road traffic speed plays a vital role on traffic control and guidance.The current research mainly focuses on estimating the traffic flow and traveling time,but seldom on of the prediction of traveling speed in short-term.Most of the existing approaches forecast the future traffic speed just by average the current traveling speeds of floating vehicles.However,to calculate the current traveling speed is not a straightforward task.Such approaches have to determine whether the data of floating vehicles are abnormal,and also need to consider the waiting time before intersections,which leads to the more complicated situation and poor performance.In order to avoid the difficulty of calculating the speed of a single floating car,a method of short-term traffic speed forecasting is proposed in this paper,which takes into account the influence of traffic states of adjacent roads(such as traffic flow,traffic speed,road occupancy and traffic density)on the traffic speed,and uses the piecewise linear function as the mapping function of each traffic state and traffic speed.We use the gray relational analysis method to select the traffic state of each road which is most relevant to the traffic speed.The Jenks clustering method based on the dynamic programming method is applied to determine the segmented range of the segmented function.Finally,the particle swarm algorithm is used to solve the function parameters.In addition,in order to determine the exact location of the floating vehicle on the road and reduce the influence of GPS sampling point location error,we design a map matching algorithm based on grid structure.The matching algorithm takes into account the position,orientation and shape similarity of the corresponding GPS links in the trajectory,and thus improves the accuracy of map matching.In order to speed up the speed of map matching,this paper uses the grid structure to quickly obtain the candidate path,and add several pruning methods in the solution path.We use the real data to verify the two algorithms proposed in this paper.In the map matching,we use the Beijing road network map,and five vehicle trajectories to verify our map matching algorithm,compared with the other method,our method to ensure the speed of matching under the premise,can get better accuracy.In the short-term traffic speed prediction,we validate our approach based on the real datacollected from Wenzhou City,a large city located in eastern China,which include37,884,970 GPS records captured from 3,743 vehicles on 43,038 roads from February1 to 3,2015.The extensive experimental results show that,compared with the state-of-the-art approaches,our approach has the higher stability and accuracy,especially for 5-minute and 10-minute speed prediction. |
