This paper guides transport planners in making the best use of mobile phone traces, derived either from mobile network data or from smartphone app data. It suggests combining such new data sources with conventional travel surveys whose sample size and cost could ultimately be reduced. In the context of a rapidly evolving mobility landscape, with new modes and new services available, big data can help monitor behaviour change, learn from quasi-experiments and develop next-generation travel demand modelling tools.

This paper attempts to draw some lessons form the economic crisis of 2008-2009 in respect of transport demand modelling. It looks in the following issues: the poor treatment of uncertainty and risk, the limitations of our best theories of human behaviour, the emphasis in equilibrium modelling and the questions whether more complex and disaggregate models are more accurate tools for forecasting.

It is accepted that one of the most significant problems generated by congestion is not only the increase in travel time itself, but also a deterioration of its reliability. Drivers often complain that it is the unpredictability of travel time that they dislike most of a journey in a congested city or area. From the point of view of demand, surveys can be undertaken to ascribe values to changes in the reliability of travel time, probably using Stated Preference, or related techniques. This can contribute to the development of richer and better utility functions for use in aggregate or disaggregate choice modelling. However, in their application, these models will require predictions of how the reliability of travel time will change when conditions in the network change: these are the supply effects. One would expect reliability to depend on similar factors to those affecting congestion: capacity, travel time and flows. However, there may be also some new factors like the number of junctions encountered in a route, their type and so on. This chapter reports on an exploratory research project aiming at developing simple and manageable models of these supply effects. The models were developed using a combination of simulation and survey work, one complementing the other. The resulting models seem to explain a good deal of the variability of travel time encountered in the study area. Nevertheless, further work is needed in other areas and with other functional forms to develop more transferable and robust models.

It is accepted that one of the most significant problems generated by congestion is not only the increase in travel time itself, but also a deterioration of its reliability. Drivers often complain that it is the unpredictability of travel time that they dislike most of a journey in a congested city or area. From the point of view of demand, surveys can be undertaken to ascribe values to changes in the reliability of travel rime, probably using Stated Preference, or related techniques. This can contribute to the development of richer and better utility functions for use in aggregate or disaggregate choice modelling. However, in their application, these models will require predictions of how the reliability of travel time will change when conditions in the network change: these are the supply effects. One would expect reliability to depend on similar factors to those affecting congestion: capacity, travel time and flows, However, there may be also some new factors like the number of junctions encountered in a route, their type and so on. This chapter reports on a exploratory research project aiming at developing simple and manageable models of these supply effects. The models were developed using a combination of simulation and survey work, one complementing the other. The resulting models seem to explain a good deal of the variability of travel time encountered in the study area. Nevertheless, further work is needed in other areas and with other functional forms to develop more transferable and robust models.

During 1992 and 1993 a study was carried out in Santiago to evaluate the SCOOT system of traffic control against the fixed time plans obtained with TRANSYT over an experimental network of 34 traffic signals. This paper discusses the main methodologies considered in the study, the results obtained and the profitability indicators produced. Journey times for different vehicles and time periods are compared on 3 types of sub-network using TRANSYT and SCOOT. SCOOT was found to produce important benefits to networks while the number of public transport vehicles was low. For the covering abstract see IRRD 875078.

This paper reports on part of the Irish input to the TRENEN project established under the JOULE II Research Programme of the EU on non-nuclear energy, where the main objectives of the Civil, Structural and Environmental Engineering Department of Trinity College Dublin (TDC) and Steer Davies Gleave (SDG) team were: a) to establish the best functional form for a flow delay relationship for input to the urban optimisation model (TRENEN) using the SATURN-SATCHMO model of Dublin compiled during the Dublin Transportation Initiative (Steer Davies Gleave, 1994), b) to calibrate the TRENEN model for Dublin, and c) to examine methods of combining the use of an aggregate economics model such as TRENEN with a conventional transport network model to address the optimal combination of price and regulatory policies in the energy-environment and transport domain. The part of the work on which this paper reports is the calibration for Dublin and one of the case studies conducted using a combination of the two urban models. As the models are, by their nature, quite different in approach, direct comparisons are difficult but one means of possible comparison is by examination of modal split which is a direct output of both models. (A) For the covering abstract see IRRD 880168.

This paper presents a freight demand estimation model which uses traffic counts. In this model, the tonnage moved between points per unit of time is represented in the form of an origin destination (O-D) matrix. In the estimation, it is assumed that the commodity movements are represented by a gravity model. Two types of gravity model are used: the GR (Gravity Model) and the GO (Gravity-Opportunity model). In the procedure, the link flows are expressed as a function of the O-D matrix. The parameters of the postulated model are then estimated so that the errors between the estimated and the observed link flows are minimised. Three methods of estimation are demonstrated: a) the Non-Linear- Least- Squares Estimation Method (NLLS), and b) two Maximum- Likelihood Estimation Methods. Two goodness-of-fit (GOF) statistical tests are used to ascertain how well the calibrated model reproduces the observed O-D matrix (the Root Mean Square Error (RMSE) and relative %RMSE, and the Coefficient of determination). An inter city freight movement data survey in Bali Province of Indonesia was used to test these models and estimation methods. In this the freight was classified into 5 commodity groups. There were no traffic counts available, but 30 were generated by loading the observed matrices onto the network using all or nothing asignment. Conclusions are provided about the best model to use in Bali, the relative accuracy of the models and what parameters this depends upon, and the number of traffic counts required. For the covering abstract of the publication see IRRD 850746.

The basis of much current transport modelling for London is the four stage LTS model, run by the MVA consultancy for the Department of Transport, and implemented on a large mainframe computer. This is a sophisticated and relatively detailed model. It is doubly constrained, and has an iterative structure with mode split preceding distribution. It provides a common framework for planning by different authorities, but is somewhat cumbersome and slow to run. London Transport has previously developed a micro-based public transport assignment model - RAILPLAN version 1 - in the EMME/2 transportation planning package, using Public Transport demand matrices from LTS. The paper reports work to extend this assignment model, in order to include incremental distribution and public/private model split models. These incremental models are also based on LTS. They give the facility for estimating the effects of rail schemes on trip distribution, public/private mode split, and highway congestion within a convenient and flexible micro-based environment, building on the existing RAILPLAN assignment model. The paper concentrates on methodological issues, and so only gives a limited number of numerical results. (A) For the covering abstract see IRRD 860299.