Motivation

The FCD systems implementation can be distributed into two primary methods.

One is uses moving vehicles equipped with GPS receivers and wireless communications are used as probes to measure the road traffic information [11-16].

The data collected from GPS-equipped probe vehicles may include: location, timestamp, speed, and heading. As the electronic technologies advance, GPS receivers with wireless communication capability are getting cheaper and cheaper. In addition, because of the widely use of smart phones with GPS capabilities, Mohan et al. [8] has proposed a method using mobile smartphones as probes to monitor the traffic conditions. So the number of probes can be huge.

3

Another form of FCD systems use cellular network control messages, such as location update and handover, to detect traffic speed and congestion [9]. One of the advantages of using cellular-network control message to estimate traffic information is that it requires little extra cost, since, since cellular networks have already been deployed and mobile phones on moving vehicles are used as probes. Compared with the conventional GPS-equipped probe vehicles, this approach does not require any additional on-vehicle devices, and there are sufficient probes since MSs are so pervasively used. However, the accuracy of the traffic speeds estimated by this technique is lower than that obtained from GPS-equipped probes. To estimate traffic speed using cellular network control messages, we need to locate a mobile phone at two different locations and divide the distance of the two locations by the time elapsed. The errors in locating a mobile phone would result in errors in estimating the traffic speed. However, a cellular network can’t pinpoint the exact location of a mobile phone from the control messages exchanged between the mobile phone and the network. For example, from a normal location update (NLU) message, the network can only locate a mobile phone in an area of a cell, whose diameter is typically hundreds of meters. Therefore, NLUs can only be used to estimate the traffic speed of road segments of length dozen of kilometers, but not for short road segments.

To deal with this problem, handover messages are used to short road segments.

However, handovers between two neighbor cell also occurs in an area between the two cells, which means errors in location a mobile phone performing a handover In addition, the cell breathing technique that allows overloaded cell to reduce its size also increases the errors in locating an handover event. Since the location of a mobile phone can only be roughly located, it is difficulty identify the route that a mobile phone travels from the handover events that the mobile phone performs.

4

On the other hand, the position errors in locating GPS-equipped probe vehicles are typically within tens of meters. Therefore, traffic speed obtained GPS-equipped probe vehicles are much more accurate than that from cellular control messages.

Traditionally, fleet vehicles, such as taxi and public transportation buses, are equipped with GPS receivers and used as probes. With smart phones equipped with GPS receivers so widely used nowadays, it is feasible to used smart phones as probes. In thesis, we only consider the GPS approach.

Collecting traffic information from GPS-equipped probes can be carried out in a centralized or decentralized structure. In a centralized structure, a centralized server called Traffic Information Center (TIC) collects traffic data from a group of Probe Cars (PCs). The PCs measure the traffic data of the road segment where the PCs currently travel, and report the measured data to the TIC periodically or when the PCs passes predetermined locations. From the traffic data collected, the TIC generates traffic speed and/or travel time for each road segment. If a road segment has not been traveled by probes for a period of time, the TIC may use historic traffic information.

The traffic information generated by the TIC can be delivered to PCs and general vehicle drivers in two ways: pull or push. In “pull” approach, PCs can send requests to TIC and retrieve the traffic information by wireless communication. In “push”

approach, the TIC broadcasts the generated traffic information periodically. In order to get the accurate traffic condition of the whole road network, PCs have to be distributed widely and evenly. To achieve this, smart phone users in vehicles can be used as probes. The reason is, first, most smart phones have built-in GPS receiver.

The position information not only includes the position information but also has speed and heading information. Second, the mobile networks have been widely built. Using smartphones as probes, the volume of probe would be huge and the probes would be

5

widely spread. Most traffic information systems we have studied adopted this type of structure. In this thesis, we also use the centralized structure to design our system.

Wischhof et al. [10] have proposed a decentralized traffic information system;

PCs exchange traffic data based on inter-vehicle communications. Each PC broadcasts traffic data to other probe cars periodically; no central server exists in their system.

Therefore, this type of structure can be viewed as a peer-to-peer (P2P) architecture.

The advantage of decentralized approach, it does not need a server to collect the traffic information, so there is no bottleneck problem. But the wireless broadcast system infrastructure has not been fully built completely.