| With the gradual development of China's economy and society,the amount of motor vehicl in large and medium-sized cities has grown sharply in recent years.Along with environmental pollution,traffic accidents,and especially traffic congestion problems are becoming increasingly serious.The measures of solving urban traffic problems in recent years mainly focusing on restricting the growth of vehicle ownership and vigorously developing public transport facilities,but these measures are not the fundamental way to solve urban traffic problems.The development of intelligent transportation systems is one of the effective methods to solve urban road traffic congestion,environmental pollution and other issues.Short-term travel time prediction of urban roads is an important research field in intelligent transportation system,whitch can achieve traffic induction and effectively alleviate traffic congestion in urban roads.However,at present,the travel time prediction problems at home and abroad are mostly concentrated on data based on GPS floating cars and fixed coil detectors,and few RFID electronic license plate data are used for related research.In recent years,with the vigorous investment in the research and application of intelligent traffic systems in China,the deployment of traffic flow parameters automatic acquisition devices in most cities has been realized.Taking Chongqing as an example,it has taken the lead in popularizing the use of RFID electronic license plates throughout the country,and has deployed RFID electronic license plate collection points in major urban roads in Chongqing,which provides a rich source of data for urban road traffic research.Based on the RFID electronic license plate data in Chongqing,the main research contents in this article can be summarized as follows:(1)Based on the Chongqing RFID electronic license plate data,the related characteristics of urban road travel time were studied.Firstly,the method of obtaining the travel time of a section through RFID electronic license plate data is described.According to urban roads can be divided into fast roads,trunk roads,secondary roads and branch way,the distribution characteristics of different road trip times are analyzed,and Gaussian distribution and gamma distribution are used to fit the travel time probability distribution,and the fitting effect was analyzed by calculating the SSE and R-Square values.The results show that the lognormal distribution and the gamma distribution can be better used for fitting the travel time probability of urban road sections.(2)A travel time prediction model for urban roads based on KNN algorithm is proposed.Then based on the Chongqing RFID electronic license plate data,the predication model was used for experiments.The experimental results show that the average relative error percentage of the travel time prediction for fast roads and trunk roads is about 6%.What's more,it's found that the model is obviously better than the historical average model and the ARMA model through comparative experiments.The result is that the urban road travel time prediction model based on KNN algorithm can be better applied to travel time urban road prediction.(3)A real-time travel time prediction system based on Spark Streaming was implemented.In combination with Kafka and Redis,Spark Streaming is used to process RFID electronic license plate data in real time,and the KNN algorithm is run on Spark to predict travel time in real time.Finally,based on the RFID electronic license plate data of Chongqing City,a series of system tests were carried out,and the system performance analysis was carried out.The results show that the system can meet the design requirements. |
PDF Full Text Request