Increasing trips of heavy vehicles on traffic transportation environment had been causing serious congestion and air pollutions many years ago. Many traffic experts have been attempting to alleviate such problems and to minimize the social cost by developing some traffic researches and proposing the intelligent transportation system (ITS). The purpose and the essence of developing ITS are to utilize advanced communication techniques, traffic control and information to achieve a convenient, economic benefits and safety traffic environment. In ITS area, there are nine research topics. Every topic has its traffic domain and plays an important role to coordinate with each others. Such as, Advance Traffic Management System (ATMS) plays a kernel position in traffic monitor and management for making the global traffic network more smooth; The objective of Advanced Traveler Information System (ATIS) is to deliver reliable and useful real-time traffic information to travelers; Commercial Vehicle Operation (CVO) topic is about cost efficiency on private company and making convenient public transportations for users, likes taxi. In this thesis, we focus on the real-time Travel Time Prediction (TTP), which provides important information for travelers or drivers to understand how long he or she might reach the destination on their pre-trip job and skip traffic jam sections.
Travel time information which can help travelers to understand the current traffic condition for saving time through the selection of travel routes in pre-trip and en-route job. Besides, accurate travel time estimation could avoid congested sections to reduce transport costs and increase the service quality of commercial delivery by delivering goods within required time. For traffic managers, travel time information is
an important index of traffic system operation. Furthermore, using travel time information can scatter the condensed traffic volume and sharply reduce the habitual traffic congestion in effective, because people might choose various public transportations as their wishes. So, real-time TTP is a meaningful traffic index to be referred.
However, TTP is highly stochastic and time-dependent due to random fluctuations in travel demands, interruptions caused by traffic control devices, incidents, road construction, and weather conditions. In other words, TTP is affected by a range of traffic factors including speed, traffic volume, routing path selected, occupancy of road, and traffic facilities (e.g. Roads, lights, signs) as well as non-traffic factors including traffic event, weather, road construction, etc. But, most previous researches predict travel time based on the assumption of some historical or real-time traffic factors, such as speed or occupancy, and ignore the non-traffic factors. Thus, the results in the previous works may work well only in some special condition, but not in real-time traffic condition.
Study by Iryo [7] has found that level of reduction in congestion depends on the complexity of the road network. While vehicular flows on freeways are often treated as uninterrupted flows, flows on urban network are conceivably much more complicated since vehicles traveling on urban network are subject not only to queuing delays but also to signal delays. Besides, TTP for urban network has the routing problem to suggest a path on a given O,D pair as request. Hence, in this thesis, we are concerned with predicting travel time in an urban network instead of predicting in freeway or single arterial road. Many models had been proposed for travel time prediction in these decades, but most of them focus on the predicting the travel time on freeway [10,14,22] or simple arterial network [11,15,24]. Travel time prediction for urban network in real-time is hard
to achieve for four reasons: complexity of routing problem in road network, spatiotemporal data coverage problem of static traffic probing tools, unavailability of real-time sensor data, and improper precision of lacking real-time events consideration.
In the traditional way, traffic statuses are collected by loop sensors or monitored by supervising cameras, which are installed on intersections called sensor-based and site-based [10]. Then, traffic center managers analyze the collected data and discover the traffic patterns in order to make some actions for optimizing the global traffic network.
In recent years, some ITS projects use specially designated OBU installed on limited probing vehicles to collect traffic information, which called vehicle-based. However, all these methods can only get traffic information on the fixed location. Because they have cost down incentive to establish the stationary traffic detection equipments and quantity shortage in the amount of designated probing vehicles, which are not enough for covering all target traffic network in both spatial and temporal aspects. Thus, most traffic probing studies were resulted in simulated experiments. This thesis, we propose a more cost-effective traffic information collection method using location based service (LBS), which is generally described as a mobile information service is to provide useful location aware information, at a minimum cost and resources, to its user. In this method, we regard the vehicles of LBS-based applications as the traffic status probing vehicles.
A vehicle of the LBS-based application is equipped with an OBU (On-Board Unit), which has GPS (Global Positioning System) positioning module and GPRS communication module. OBU collects vehicle position, traveling direction, and speed from the GPS module and uplinks the vehicle status to the backend system through GPRS module. Using the LBS-based probing vehicle is possible to collect various traffic information and concerns much larger traffic area than traditional sited-based or
TTP can be estimated from historical data by analyzing the collected traffic information from different methods as discussed above. For instance, traffic speed and location of probe vehicle can be used to compute the historical travel time. And various techniques such as AI, statistics, and mathematical, could be adopted to develop travel time estimation model. However, there are many interference factors and attributive parameters to impact the accuracy of TTP. For example: construction, accident event, and weather can influence TTP on some links.
The objective of this thesis is to propose a real-time TTP expert system on urban network which predicts travel time by linear combination of real-time and historical travel time predictors based on the request of an origin (O) and destination (D) pair. The model of this system is knowledge based mechanism, which can handle the issues of non-traffic factors, and having no cost problem of vehicle-based TTP as well as the coverage problem of site-based or sensor-based TTP. We utilize the raw data of location based services (LBS); transform it into the traffic information by combining the geographical information system (GIS), then use data mining technique (A traffic pattern mining system-TPMS) to find some significantly historical traffic rules and patterns in various traffic conditions, and predict travel time by integrating these historical traffic patterns, real-time traffic information and real-time external information sources. The external information sources provide the real-time information may affect TTP, but meta-rules offered by traffic experts dynamically tune the combination weights of historical and real-time TTP in order to raise the precision of real-time TTP. For example, if a current car accident is happening on the link of OD pair, TTP system may trigger one of the meta-rules to raise the weight of real-time TTP on that link. Because the delay of travel time on that link will be reflected immediately by the real-time LBS, thus raise the weight of real-time TTP will get higher precision.
The rest of this thesis is organized as follows. Chapter 2 shows the related works of traffic probing tools and TTP issues. Chapter 3 gives the introduction of LBS, and talks about the important historical traffic information, which is derived from LBS. And Chapter 4 is the kernel part in this thesis, which describes our target real-time TTP expert system and separates four phases to organize our system. Each phase goes detail in each section. In Chapter 5, we implement the prototype of TTP system in Taipei urban network, and utilize the taxi dispatch system as our LBS data source. Real-time, historical and linear combination predictor are evaluated and compared in this chapter.
Finally, conclusions and future research are presented in Chapter 6.