The Estimation Of Real-Time Traffic State Based On Floating Car Data

Real-time traffic state is the basis of the Advanced Traffic Information Systems (ATIS). It can help to implement the traffic control and management, dynamic route guidance, improve the travel condition and provide the reference for resident trips, bus dispatching and urban planning.Traditional traffic detector (such as loop detector, microwave detector, radar detector) plays an important role in the detection of traffic flow. But there are many drawbacks in the detection. For example, the traveling time and road speed is not available, the road space to cover by detectors is limited, and the cost of installation and service is too high. In recent years, the floating car technology is widely applied to collect traffic information, especially the vehicle terminal based on GPS is installed on the floating cars, like buses, taxis and logistic vehicles. These floating cars running on the urban road network provide a number of real time traffic data, such as vehicle location, vehicle speed, vehicle condition, etc. The floating car data has many advantages, such as high precision, large amount of data, wide space of coverage, economy and convenience.Single type vehicles are selected ad floating cars by most of domestic and foreign research. But the single type vehicles cannot represent the constitute of entire road traffic flow. And these studies are based on many assumptions of long period, large sample size, high sampling frequency. But in the actual situation, these assumptions are difficult to meet, so the accuracy of the estimation value is low.This thesis choose the service-oriented vehicles for public transport (buses and taxis), and logistics vehicles as floating cars to collect vehicle position, vehicle speed and other vehicle information by many sensors equipped on vehicles like GPS. The GPS positioning accuracy is affected by many factors, after data preprocessing, vehicles can be matched on the road by using of map matching algorithm. Given the number of floating car, the distribution of vehicles in the road network is nonuniform. So the number of vehicles on each road can not meet the requirement of minimum sample size. Meanwhile traffic state is influenced by the length of road section. This thesis proposes a reasonable method to divide road sections to reduce the impact of signal delay and delay in bus stops, and integrate road sections to one section dynamically to overcome the drawbacks of low number of vehicles on the road. By Vissim simulation, it is founded that the average speed of single-type floating cars has the same changing trends with the one of cars in the same road section, so the least square method is used to fit this speed of single-type floating cars to car speed. After getting the fitting speed of three types, and then a new weighted mean algorithm is applied to get the average speed of road section. Simulation for three times is used to create data to modify the model proposed by Vissim software. The first-time simulation provides data for the least squares curve fitting algorithm. The second-time and third -time simulation provide data to validate the accuracy of model.Through the above two simulation tests, the average speed estimation model based on the data of multi-types floating cars proposed in this thesis can effectively limited the average error rate at less than 7%, the absolute error between the average speed estimated and the true value was less than 5km/h on 92% of roads. The simulation experiment results show that the model is reasonable and the algorithm is effective.