Antony Stathopoulos

Fuzzy Modeling Approach for Combined Forecasting of Urban Traffic Flow

This paper looks at the problem of accuracy of short-term traffic flow forecasting in the complex case of urban signalized arterial networks. A new, artificial intelligence-based approach is offered for improving accuracy of traffic predictions through suitably combining forecasts derived from a set of individual predictors. This approach employs a fuzzy rule-based system (FRBS), which is augmented with an appropriate metaheuristic (direct search) technique to automate the tuning of the system parameters within an online adaptive rolling horizon framework. The proposed hybrid FRBS is used to nonlinearly combine traffic flow forecasts resulting from an online adaptive Kalman filter and an artificial neural network model. Empirical results obtained from the models implementation into an actual urban signalized arterial show the ability of the proposed approach to considerably overperform the given individual traffic predictors.

Real-Time Traffic Volatility Forecasting in Urban Arterial Networks

A methodology is presented for forecasting traffic volatility in urban arterial networks with real-time traffic flow information. This methodology provides a generalization of the standard modeling approach, in which both the mean, modeled by an autoregressive moving average process, and the variance, modeled by an autoregressive conditional heteroscedastic process, are time-varying. The statistical analysis and forecasting performance of the proposed model are investigated with real-time traffic detector data from a real urban arterial network. The results indicate the potential of the proposed model to improve the accuracy of predicted traffic volatility across different lengths of forecasting horizon in comparison with the standard generalized autoregressive conditional heteroscedastic methodology.

Reliable stochastic design of road network systems

This paper investigates the continuous version of the stochastic Network Design Problem (NDP) with reliability requirements. The problem is considered as a two-stage Stackelberg game with complete information and is formulated as a stochastic bi-level programming problem, which is extended to include reliability as well as physical and budget constraints. The estimation procedure combines the use of Monte Carlo simulation for modelling the stochastic nature of the system variables with a Genetic Algorithm (GA), for treating the complexity of this new formulation. The computational experience obtained from a test road network application demonstrates the ability of the proposed methodology to address the need for incorporating reliability requirements and stochasticity into the various system components in the design process. The results can provide useful insight into the evaluation of alternative reliable network capacity improvement plans under the effect of uncertainty on the demand, supply and route choice process of travellers.

Real-Time Dynamic Origin-Destination Matrix Adjustment with Simulated and Actual Link Flows in Urban Networks

The efficiency and robustness of different real-time dynamic origin–destination (O-D) matrix adjustment algorithms were investigated when implemented in large-scale transportation networks. The proposed algorithms produce time-dependent O-D trip matrices based on the maximum-entropy trip departure times with simulated and actual observed link flows. Implementation of the algorithms, which are coupled with a quasi-dynamic traffic assignment model, indicated their convergent behavior and their potential for handling realistic urban-scale network problems in terms of both accuracy and computational time. The main factors influencing the numerical performance of each algorithm were identified and analyzed. Their relative efficiency was found to be particularly dependent on the level at which the assigned flows approximate the observed link flows. These results may provide insights into the suitability of each algorithm for diverse application domains, including freeways, small networks, and large-scale urban networks, where a different quality of O-D information is usually available.

Genetic computation of road network design and pricing Stackelberg games with multi-class users

This paper deals with the problems of optimal capacity and pricing decisions in private road networks. These problems are described as a class of design and pricing Stackelberg games and formulated as nonconvex, bilevel nonlinear programs. Such games capture interactions among the decisions of system designer/operator, government regulations and reactions of multi-class users on optimal toll-capacity combinations. The present class of games applies to a realistic urban highway with untolled alternative arterial links. In contrast with the mostly used continuous representations, the highway capacity is more intuitively expressed as a discrete variable, which further complicates the solution procedure. Hence, an evolutionary computing approach is employed to provide a stochastic global search of the optimal toll and capacity choices. The results offer valuable insights into how investment and pricing strategies can be deployed in regulated private road networks.

Short-Term Prediction of Urban Traffic Variability: Stochastic Volatility Modeling Approach

This paper addresses the problem of modeling and predicting urban traffic flow variability, which involves considerable implications for the deployment of dynamic transportation management systems. Traffic variability is described in terms of a volatility metric, i.e., the conditional variance of traffic flow level, as a latent stochastic (low-order Markov) process. A discrete-time parametric stochastic model, referred to as stochastic volatility (SV) model is employed to provide short-term adaptive forecasts of traffic (speed) variability by using real-time detector measurements of volumes and occupancies in an urban arterial. The predictive performance of the SV model is compared to that of the generalized autoregressive conditional heteroscedasticity (GARCH) model, which has been recently used for the traffic variability forecasting, with regard to different measurement locations, forms of data input, lengths of forecasting horizon and performance measures. The results indicate the potential of the SV model to produce out-of-sample forecasts of speed variability with significantly higher accuracy, in comparison to the GARCH model.

Genetic Algorithm-Based Approach for Optimal Location of Transit Repair Vehicles on a Large Urban Network

Determining the optimal location of a fleet of vehicles is necessary in a number of potential applications, such as special repair vehicles for buses on a large public transportation network. The Athens Urban Transport Authority operates a large bus fleet over an extensive network for 19 h a day and serves a population of approximately 4 million people, all in a heavily congested road network. During the 2004 Summer Olympic Games, held in Athens, most spectators, employees, and volunteers were transported to and from Olympic Games venues by public transportation. Dedicated Olympic Games bus lines operated under a tight around-the-clock schedule. During normal operations and particularly during events such as the Olympic Games, incidents such as vehicle breakdowns and minor accidents can have a severe effect on the operation of the public transport network and can cause a significant decrease in the level of service. To help the authority locate bus repair vehicles over the entire network, a decision support system was developed on the basis of an embedded genetic algorithm used for obtaining optimal location solutions. The systems design and performance make it easy to operate under real-time conditions, which is useful for planning and for fast vehicle redeployment.

An agent-based framework for designing competitive transportation systems

This paper deals with the case of programming the development of future transportation systems by identifying inter-dependencies among competitors. Here, a market of maritime facilities is modelled as an n-person non-cooperative game among port authorities that promote the attractiveness of their terminal facilities, in terms of level of service provided. At the same time, freight shippers/carriers who are forming their service network based on the prevailing conditions offered by the available transportation paths are also modelled in the context of a non-cooperative game. Optimal decisions are obtained by extending the standard single leader–multiple followers Stackelberg game-theoretic formulation of the network design problem (NDP) to its complete form of multiple leaders–multiple followers competitive NDP. The estimation of the equilibrium point of the above complex transportation system is based on a novel evolutionary optimisation framework. Results from alternative design strategies are presented revealing the effects of competition and cooperation on systems design.

Fuzzy Rule-Based System Approach to Combining Traffic Count Forecasts

Current advances in artificial intelligence are providing new opportunities for utilizing the enormous amount of data available in contemporary urban road surveillance systems. Several approaches, methodologies, and techniques have been presented for analyzing and forecasting traffic counts because such information has been identified as vital for the deployment of advanced transportation management and information systems. In this paper, a meta-analysis framework is presented for improving forecasted information of traffic counts, based on an adaptive data processing scheme. In particular, a framework for combining traffic count forecasts within a Mamdani-type fuzzy adaptive optimal control scheme is presented and analyzed. The proposed methodology treats the uncertainty pertaining to such circumstances by augmenting qualitative information of future traffic flow states (and values) with a knowledge base and a heuristic optimization routine that provides dynamic training capabilities, resulting in an efficient real-time forecasting mechanism. Results from the application of the proposed framework on data acquired from realistic signalized urban network data (of Athens, Greece) and for a diversity of locations exhibit its potential.

Simultaneous Origin-Destination Matrix Estimation in Dynamic Traffic Networks with Evolutionary Computing

This paper presents an evolutionary computing approach for the estimation of dynamic Origin-Destination (O-D) trip matrices from automatic traffic counts in urban networks. A multi-objective, simultaneous optimization problem is formulated to obtain a mutually consistent solution between the resulting O-D matrix and the path/link flow loading pattern. A genetically augmented microscopic simulation procedure is used to determine the path flow pattern between each O-D pair by estimating the set of turning proportions at each intersection. The proposed approach circumvents the restrictions associated with employing a user-optimal Dynamic Traffic Assignment (DTA) procedure and provides a stochastic global search of the optimal O-D trip and turning flow distributions. The application of the model into a real arterial street sub-network demonstrates its ability to provide results of satisfactory accuracy within fast computing speeds and, hence, its potential usefulness to support the deployment of dynamic urban traffic management systems.