In this chapter, we review papers related to travel time estimateion methods.
Section 2.1 includes papers in corresponding with estimateion travel time on highway, while Section 2.2 is about that in urban area. Section 2.3 summarizes these reviews and makes a conclusion.
2.1 Review of the Literatures relating to estimateion travel time on highway
Most of the papers about travel time estimateion on highway are based on traffic flow theory and artificial neural network to estimate travel time.
Jasperse, Peterson, and Yamance use flow theory to estimate vehicle travel time, while Lee and Rong use artificial neural network. Concepts of these estimateion methods are explained as follows.
In order to provide drivers with useful and correct information, Jasperse [2]
develop algorithms to calculate travel time and queue-length. Main source of the input data is from loop detector (vehicle speed and traffic flow). The algorithms are proved useful by comparing the outcome of the algorithms with real data recorded by another system. Also the algorithms will be briefly outlined. The findings are that in the common cases (short and long-closed road-segments ) the estimation of travel time is satisfactory. The queue-length estimator for long closed road-segments seems also adequate. The algorithms doesn’t function for open-long segments and with large absences of the input data.
Paterson [3] presents a new macroscopic algorithm for estimateing freeway
travel time. The proposed algorithms has the ability to be used in any traffic situation between two points. It also has the ability account for relative speeds and densities of vehicles within the traffic stream. Calibration and field testing of the new algorithm of has indicated that substantial improvements in travel time estimateion can be achieved when compare to existing system currently operational in Melbourne, Australia.
Yamance [4] develop the Travel Time Estimation System to estimate the travel tome using the data of Automatic Vehicle Identification (AVI) and ultrasonic vehicle detectors (UVDs). This system collects technique and estimates travel time for 106 major sections, taking account of data from UVDs. The key technology of the system is being able to estimate travel time accurately in various traffic situations. They develop a new algorithm for travel time estimation. In the algorithm, if the estimated travel time is considerably long, drivers may experience delays in receiving travel time information that may lead to incongruities with actual travel time. Therefore, they develop two more methods to overcome this problem.
Lee [7] uses artificial neural networks to build a travel time forecasting model. The input data is the real time traffic information. The traffic information involves the GPS data of bus, the data of vehicle detector and the data of incident. The section of highway in this research is from Xi-luo to Yong-kang interchange. It was divided into several segments for model building. The effects of four separating ways to the travel time forecasting are discussed fewer data sources. This study showed very promising practical applicability of the proposed models in the Intelligent Transportation System ( ITS) context.
traffic information in first phase. Second, he uses artificial neural network with three layers, fully connected and feed-forward, and backpropagation algorithm to build forecasting highway travel time models. Using simulation to produce the relative data of traffic character detected by vehicle detector sequentially for the base of artificial neural network training and testing. The result of this research can be provided to forecast travel time in real time highway travel time estimation.
2.2 Review of the Literatures relating to estimateion travel time on urban area
Most of the papers about travel time estimateion in urban area are based on traffic flow theory and mathematics models to estimate travel time.
Sanghoon uses flow theory to estimate vehicle travel time, while Chang and Wu used mathematics algorithm. Concepts of these estimateion methods are as follows.
Sanghoon Bae [10] uses Automatic Vehicle Location (AVL) system equipped buses as probe vehicles to estimate bus travel time. Since many transit organizations throughout the North America are currently operating these AVL buses on their bus routes, in a sense, existing AVL bus would be the most cost-effective probe vehicle that can be utilized for data collection in the proactive mode. The initial objective for the first goal was to model and simulate the dynamic bus behaviors at a single and multiple buses using time varying passenger arrival rate and passenger boarding rate.
Then, a prototype arrival time estimation model is simulated by adopting the Parameter Adaptation Algorithm (real-time identification). Later in the research, a dynamic link travel time function was developed by dividing the conventional arterial
link into two regions.
Kuo [15] divided the estimateion model for bus arrival time into the long-term and short-term models by time lag to fulfill the different demand of passengers. In the building of long-term model, this paper applies dynamic and stochastic travel time to calculate the expectation and variance of link travel time and uses Fast Fourier Transform (FFT) to obtain the function types. In the other hand, this study proposes three short-term estimateion models and evaluates them by precision, robustness and stability After model building, this study finds that the point of demarcation of long term and short-term estimateion models is 35 minute. In validation, it discusses the estimateion error with respect to route distance, time lag and intersections. At last, this study also validates the model precision on the multi-route through reality investigation.
Wu [8] used GPS positioning information to estimate vehicle travel time. In this research mainly using vehicle historical travel information to estimate vehicle travel time. In order to make the estimation module to fit both the interurban and urban trips, he divided the vehicle travel time into vehicle running time and vehicle delay time. At the same time, in order to amend the shortcoming that it could not response the sudden change of vehicle running pattern of using historical travel information to estimate vehicle travel time. He use the GPS positioning information, includes vehicle average running speed and former vehicle travel time information to adjust the vehicle running time estimation. And he used the vehicle real stopping behavior to adjust the vehicle delay time estimation.
Chang [9] proposes a data processing model to process the real-time travel data
link. However the buses have their main missions to do, such as carrying passengers, we must filter out the unneeded data to provide accurate travel information. The data processing model includes two parts. One is the data filtering model to filter out the un-normal low speeds resulted from stops for collecting passengers or red light. The other is the data-cutting model. This model using the change point analysis of statistic theory to find a cutting point where the data has significant difference. The result showed that the data filtering model and the data-cutting model could work well.
2.3 Summary
1. I this study, we choose to focus our research on travel time estimation in urban area. The reason is that, although we found that most papers are mainly written for that on highway but rarely for urban area, the
importance of urban area travel time estimation in APTS, however, is still none the less.
2. This study takes GPS as its main data source for the sake of cost and labor.
A majority of papers are based on traffic flow theory and artificial neural network. But through those methods, we have to install a lot of detectors in urban area, and that costs much.
3. Former researches have done well in processing travel time and historical vehicle speed data. So, by using those reviewed results, this paper is intended as a further research in waiting time estimateion, hoping for estimateing bus traveling time more accurately.