ABSTRACT

After introducing the principle of float car data (FCD), this paper gives the primary flow of pre-handing and mapmatching of the FCD. After analyzing the percentage of coverage of FCD on the road network, large quantity of heritage database of routing status is used to estimate the routing velocity when lack of FCD on parts road segments. Multi liner regression model is then put forwarded by considering the spatial correlativity among the road network, and some model parameters are deduced when time series is classified in day and week. Besides, error of velocity probability and error of status probability are achieved based on the result from field testing while the feasibility and reliability of the velocity estimation model is obtained as well. Finally, as a case study in Shanghai center area, the whole routing velocity in the road network is estimated and published in real time.

1. Introduction

Efficient traffic control and reliable traffic information services require reliable status information of entire road network in real time. Generally, traffic status can be detected by fixed sensors, such as inductive loops, video systems [1]. However, vehicle speed and traffic flow measured by fixed sensors couldn’t completely cover the entire road network. Actually, traditional observation about traffic flow in space and time has been hardly reflected complexity of traffic itself in micro and macro [2].

In recent years, since the mobile sensors are widely used in many traffic conditions identification, the floating car data method has become an instrument for estimating the routing velocity in the road network. This method uses traffic information such as car location, speed and direction gathered by a highly distributed moving vehicle in the traffic flow, to calculate the road velocity based on certain model. For example, using GPS sensors placed in cars can obtain location and velocity of a large number of vehicles; these data play an important role in traffic planning and management, as well as Intelligent Transportation System and Service. That is any GPS equipped vehicles can be a sensor to collect the real-time traffic information of the routs where vehicles drove. These data are needed for describing traffic congestion, calculating routing velocity and publishing current traffic announcement, etc [3].

As a new method of urban traffic information collection, floating car is emphasized by both domestic and foreign experts of ITSS or enterprises, and there have been a number of theories and applications about FCD [4]. ADVANCE was launched in 1991 as a major test of a dynamic, in-vehicle route guidance system in the United States. The objective was to determine if motorists would be given information which would help them avoid congestion and improve the quality of their trip. Germany Aerospace Center used 4000 sensors placed on roadways and 25000 collecting vehicles from BMWs and Volkswagens to provide service for each OEM [5]. The UK company Trafficmaster started to collect and process traffic data since it announced to be a private company in1998, and it offered traffic information service based on the data collected from static sensors and supplemented by FCD [6]. P-DRGS in Japan gathered various kinds of vehicle information, and then sent traffic predictive information to the navigation device installed in cars by using wireless communication. At the same time, navigation device installed in cars is also used to collect traffic information and realize traffic forecast. Swiss OPTIS was committed to studying the data collecting method that could provide good travel information links at reasonable cost. In comparison, intelligent traffic construction started late in china. Colleges, research institutes, enterprises and public institutions began to stat cooperation and do independent research on the floating car data collection and processing [3]. Now, the number of floating car used as data collection terminal has developed to a certain scale in some cities, such as Beijing, shanghai, Guangzhou, Hangzhou, Ningbo, etc.

The existing studies related to the traffic information collection system based on floating car mostly concentrate on the costs, floating car quantity, system structure and accuracy, etc [7]. Due to the unique mobility and limited number of floating car, the real-time floating car data could not cover all the road networks, using appropriate method could guarantee data integrity and improve efficiency of application. Therefore, exploring the coverage of floating car data and method for estimating and information matching are practically, this is also major issue in the paper.

2. Floating Car Data Collection and Map Matching

2.1. Floating Car Data

Floating car data refers to the data (such as speed, distance and travel time) being collected by sensor equipped vehicles roaming the road network in a fixed time interval. The data are sent to the information center in a variety of ways. After being analyzed, traffic data are stored in a large-scale shared database which provides varied real-time traffic information for different road users. The essence is that using sample from vehicles overall network to reflect the general road conditions.

Using GPS installed in vehicle to collect FCD enable us to obtain a series of discrete location data, its format is similar to the common GPS data format. Generally, the format of floating car data is relatively flexible and location data is required. In this paper, real-time location data is based on the existing 3000 taxis in shanghai, these taxis record information including time, longitude/latitude coordinate and instantaneous velocity with 2 seconds sampling interval. The traffic control center set 30 seconds as time interval to receive information, that is, vehicle position information is obtained every 30 seconds under normal circumstances [7]. After analyzing, the system promulgates information in 5 minutes interval, so the average speed of road segment for each 5 minutes is deduced. In accordance with the road grades, smooth, crowed and congestion are denoted by green, yellow and red, the complete city real-time traffic map could be obtained combing with map performance. The floating car data format used for this research is shown in Table 1.

2.2. Coverage of Floating Car Data in Road Network

Floating car data collection system requires a combination of map to performance thematic map, so the road network vector meet the system requirement needs processing [3]. Figure 1 shows 1 hour historical data points and partial enlargement in central shanghai based on 3000 floating car. As can be seen from Figure 1, the position points of FCD can cover the entire central network well, according to the partial enlargement; using floating car data is a good method to simulate the real condition of road network. With the distribution of historical vehicle data accumulated over a certain period of time, the percentage of coverage of FCD is sufficient for effectively describe the distribution of real road network.

2.3. Map Matching

FCD map matching is one of the key technologies of the traffic information collection and processing based on floating car, the map matching compares track of vehicle location calculated by position device with road data in electronic database, determine the most driving road segment and position in this road segment by a specific algorithm, so as to estimate routing velocity.

Map matching concepts and method mainly come from the vehicle navigation system [8]. However, the purpose of floating car map-matching, object size, speed and matching rate requirement are very different with the vehicle navigation map matching, so these techniques can not directly be applied to the floating car traffic information processing. First of all, the purpose of floating car map matching is to transform vehicle position information into travel time of the road, it doesn’t need high matching rate as vehicle navigation map matching does. Secondly, the object scale of floating car map matching is large, the processing has to complete thousands of cars matching with map and require a high matching speed.

In generally, real road status information system based on FCD receives floating car data and matches to the appropriate road segment, then estimates average route velocity according to the average speed of each vehicle driving on the road. However, 1) owing to the high mobility of floating car, the percentage of coverage of the road network may not reaches 100% in non-peak hours, so some route velocity can’t be calculated as a result of missing FCD information. 2) wireless communications probably lead to losing data packet. 3) taxis are limited in the certain period of time on some road segments. These special circumstances would restrain the using of FCD, with the general algorithm useless; route velocity information gap will appears. In order to effectively address these problems, appropriate model must be estimated to perform route velocity under missing data and release the real-time traffic information.

3. Multiple Models for Velocity Estimation in Spatial Road Network

3.1. Velocity Estimation Error

Figure 2 shows time change map of coverage arterial road in shanghai at June 2008 based on FCD. According to Figure 2, average coverage of speed data on arterial road is about 60% in daylight, to obtain complete road network speed data, velocity of roads without speed data coverage need to be estimated.

Actually, the floating car real-time traffic system provides speed data every 5 minutes, the data file is named in the format “year month date_time number.txt”, the complete data of a day contain 288 files of time. As the data file at 0:00 - 0:05 on the April 1, 2008 are named as “20080401_001.txt”. Each record in the file includes code of road segments, forward speed and reverse speed. If the road lack floating cars signal to calculate velocity, this direction velocity is given the value –1. The evaluation model in the paper uses speed data provided by shanghai real-time road status system during April 1, 2008 to May 31, 2008 based on FCD, the data in June 2008 is applied to model validation. First, error in the velocity estimation is defined as follows:

1) Velocity error Suppose velocity estimated by the model is, the actual current velocity is, then.

2) Status error With the combination of map, the real-time traffic system provides the road status information, which describes current traffic condition as congestion, crowded or smooth for user. Therefore, in the accuracy analyzing of the model, status error variable is introduced in order to publish results to users more intuitive. Setting the status simulated by velocity based on the model as, actual current status is, the different figures represent different states, such as 0—congestion, 1—crowded, 2— smooth, then.

3) Error probability The essence of real-time traffic system based on FCD is to random select from all the vehicles overall the network, using the sample to reflect the traffic condition, the processing of velocity estimation remains some errors. The accuracy analysis of model allows a certain amount of error. Generally, data error satisfies and status error satisfies, these data is considered acceptable in practical application. So introduce the error probability, it is the probability of velocity error in, and          (in), and the probability of status errorin, and.

3.2. Multiple Spatial Linear Regression

In the geographic system of multiple elements, these elements interact and interrelate. According to the characteristic of spatial location, road segment velocity is closely related to its connective road velocity.

Figure 3 shows the sketch map of multiple linear regression model about relative information of road segment, suppose the velocity Y in a direction of road segment is influenced by the velocity of K relative road segments;  the observation of group is, , , , , thus the multiple linear regression model can be described as follows: