Takamasa Iryo

Properties of dynamic user equilibrium solution: existence, uniqueness, stability, and robust solution methodology

This paper reviews studies that deal with properties of dynamic user equilibrium (DUE) solutions to understand the extent to which they have been resolved in past studies and indicate theoretical problems that remain to be addressed in future research. DUE assignment is an equilibrium-based methodology for dynamic traffic assignment. In conventional equilibrium-based static traffic assignment, the four desirable properties of solutions – existence, uniqueness, stability of solutions, and the existence of a robust solution methodology – are all guaranteed. These properties must be ensured for planners to obtain a reliable solution for equilibrium-based assignment. Meanwhile, for DUE assignment, although some of these properties have been proved, many issues regarding these properties remain unresolved. This paper reviews studies that deal with these properties to understand how far they have been investigated and discusses what is required from future research to improve the robustness of DUE assignment.

Estimation of behavioural change of railway passengers using smart card data

Smart card systems are becoming increasingly popular on a global scale, not just for purchasing general goods and services, but also for paying public transport fares. When a traveller uses a public transport smart card, the exact time of their passage through ticket gates are recorded in the smart card system database. However, these data have not yet been sufficiently studied in the field of transport research. The aims of this paper are to estimate the behaviour of railway passengers by using smart card data and to evaluate the effects of train operations. In particular, the analysis is focused on the comparison of passengers’ travel choice behaviour before and after the railway company altered the train timetable.This paper describes how the passing times of individual passengers at entrance and exit ticket gates are aggregated for a small discrete time interval. Analysis of the departure, travel, and arrival time distributions shows that passengers smoothly adjusted their travel behaviour to the new train timetable. Analysis of the passing times at origin and destination station ticket gates in combination with the train timetable makes it possible to identify which train each traveller was likely to have boarded. This paper also proposes a method to assign a passenger to a combination of trains between an origin and destination stations. The method is examined using actual smart card data.

A pedestrian model considering anticipatory behaviour for capacity evaluation

This study proposes a microscopic model of a pedestrian flow that considers the anticipation and decision processes of pedestrians interacting with each other by incorporating game theory. Modelling pedestrian flow in crowded places is essential for evaluating pedestrian facilities and their capacity. It is necessary to consider the anticipation and decision processes of pedestrians interacting with each other because pedestrians are likely to collide frequently in congested situations. This study proposes a model that utilizes the concept of Nash equilibrium. An algorithm for simulating the model is proposed, along with an algorithm for determining the best route of a pedestrian considering the existence of other pedestrians. That the proposed model is capable of reproducing the congested pedestrian flow is confirmed by numerical tests.

The posterior probability distribution of traffic flow: a new scheme for the assignment of stochastic traffic flow

This study proposes a new scheme for assigning traffic flows that aims to capture the stochastic nature of route traffic flows. We consider the route traffic flows to be random variables. The distribution of these random variables is formulated as a conditional probability distribution for a given assumption: the traffic network is in stochastic user equilibrium. From a Bayesian perspective, we treat the conditional distribution as a posterior distribution of route traffic flows, which is obtained using Bayes’ theorem. We develop a basic Metropolis–Hastings (M–H) sampling scheme, as well as a M–H within Gibbs sampling scheme, to draw samples from the posterior distribution. We estimate characteristics such as the means and variances of route traffic flows from simulated samples. The proposed model can directly output the route traffic flows, and has a highly flexible computation process.

Data Mining for Traffic Flow Analysis: Visualization Approach

Data mining has attracted considerable attention as a method that can be used to discover certain characteristics from large amounts of data. In traffic flow analysis, a large amount of traffic flow data is continuously collected and stored over several years.

On the uniqueness of equilibrated dynamic traffic flow patterns in unidirectional networks

Uniqueness of the dynamic user-equilibrium assignment is still an important issue. This paper proves uniqueness with a milder condition compared to past studies and shows another counterexample to the uniqueness. A unidirectional network, in which any node on any shortest route has a unique node potential, is introduced. Orders of vehicles are determined by this node potential so that, given any two vehicles passing through the same node, the lower potential vehicle arrives at the node before the higher potential vehicle. It is shown that, for a unidirectional network in equilibrium, the link travel times and traffic volumes of congested links are uniquely determined. Moreover, a simple non-unidirectional network having multiple equilibria is introduced. This example exhibits importance of unidirectional-network structure to prove uniqueness.

A Link-Based Stochastic Traffic Assignment Model for Travel Time Reliability Estimation

This study proposes a link-based stochastic traffic assignment model that aims to capture the stochastic nature of link traffic flow, and the output of the model is the probability distribution of link traffic flows. We consider the link traffic flow variables as random variables. The distribution of the random variables is formulated as a conditional probability distribution for a given assumption: the traffic network is in stochastic user equilibrium. The conditional probability distribution is deduced from a Bayesian theorem, referred to as posterior probability distribution. A Markov chain Monte Carlo(MCMC) method is applied to simulate samples from the posterior distribution. Characteristics such as the means and variances of link traffic flows as well as travel time reliability are estimated from the simulated samples.

Preprocessing Parallelization for the ALT-Algorithm

In this paper, we improve the preprocessing phase of the ALT algorithm through parallelization. ALT is a preprocessing-based, goal-directed speed-up technique that uses A* (A star), Landmarks and Triangle inequality which allows fast computations of shortest paths (SP) in large-scale networks. Although faster techniques such as arc-flags, SHARC, Contraction Hierarchies and Highway Hierarchies already exist, ALT is usually combined with these faster algorithms to take advantage of its goal-directed search to further reduce the SP search computation time and its search space. However, ALT relies on landmarks and optimally choosing these landmarks is NP-hard, hence, no effective solution exists. Since landmark selection relies on constructive heuristics and the current SP search speed-up is inversely proportional to landmark generation time, we propose a parallelization technique which reduces the landmark generation time significantly while increasing its effectiveness.

Estimation of Passenger's Train Choice in Railway Network with Smart Card Ticket System

近年,鉄道などの公共交通機関でICカード乗車券(交 通系ICカード)の導入が進んでいる.交通系ICカードの 乗車履歴データ(以下ICデータと呼ぶ)は,各鉄道会社で 管理されており,鉄道利用者に対するマーケティング等 への活用が期待されている. ICデータの特徴には,1ICカード(利用者)毎の利用 履歴が記録されていること,2改札通過時刻を1分単位 という詳細な時間解像度で記録していること,3長期に わたる改札通過の観測データが収集できること,4入出 場記録が完結されなければ利用者は継続利用できないた めデータの欠損が少ないことがあげられる.これらの特 徴よりICデータは,時刻毎の利用者の鉄道利用の状況の 詳細な分析や,利用者行動の長期にわたる変動を分析す るのに適したデータであるといえるだろう. これまでの鉄道利用者の行動を推定する研究では, 大都市交通センサスによるデータを用いる方法が主流で あった.しかし,センサスを用いた方法では,都市 全体の鉄道旅客流動を把握することが可能である半面, 日々の変動をとらえることはできない.一方,各車両の 応荷重の履歴データや自動改札機の通過履歴によるデー タを用いることで日々の旅客の変動を推定することも可 能になってきている.これらの手法では,個々の車 両の乗客数や駅間を移動する乗客数の日々の変動を捉え ることが可能である.しかし,これらのデータを用いた 方法でも,各利用者を識別した上で変動を捉えることは できない.これに対し,ICデータでは,利用者毎の乗降 駅と乗降時刻を知ることができる.その情報から,利用 者毎に乗車列車を推定できれば,様々な利用者層での列 車選択の特性を明らかにすることが期待できる. 本研究の目的は,ICデータと列車ダイヤを用いて,

Tracking Individual Travel Behavior Using Mobile Phone

Tracking type data collection methods using mobile phone have been developed for monitoring and analyzing individual travel behavior. Previous methods using PHS (Personal Handy phone System) assumed that the location data of base stations were known. This paper shows a location positioning method of a mobile object when the location data of base stations are not available. Instead of base stations, signal strength vectors (reference vectors) are observed at many places in a target area. A mobile object also observes a signal strength vector (target vector). The similarity index of a reference vector and a target vector is used for calculating the location position of a mobile object. This method is applied to tracking travel behavior of tourists in Kobe Kitano area.