CONCURRENT SESSION 3: Big Data, Big Models
Mobile data has been recognised as a key data source in transport planning in recent years. Its strengths and weaknesses have increasingly become understood, but perhaps never before in as difficult an environment as the whole of Greater London. Project EDMONDhas been a major exercise, involving many new techniques to deal with the challenges of data development for such a complex city.
The Strategic Analysis department at Transport for London (TfL) is responsible for maintaining and updating the highway assignment and public transport models which are used to forecast future demand in London. TfL commissioned Jacobs (and their sub-consultants Telefonica and Aecom) to update these models using mobile network data from the O2 network fused with a variety of other datasets relating to transport in London.
The paper will first describe the methodology used to process the mobile network data to identify trips where users have moved between two distinct parts of the cell network. The paper will then describe how we inferred journey purpose from analysing mobile phone data each day over a 90 day period to provide insights of home and work locations, and the identification of education trips; and how we inferred the mode of the trip from innovative statistical methods to estimate a classifier, probabilistically to allocate mode, explained by information on routeing, land-use, and various other trip attributes derived from mobile network data or other fused datasets. The results of an opt-in survey will be presented, which compared trips enumerated in a travel diary from a sample of respondents’, with the trips inferred from the data processed, with permission, from their mobile phones.
The paper will then describe some of the insights that we have been able to derive from the output matrices, including trip patterns not previously modelled, e.g. weekends, night-times and of tourist visitors to London, and some of the further in-depth analysis than could be undertaken on the database.
Big data is often talked about in the abstract. We can now collect huge numbers and apply clever algorithms, so we think the answers must always be better. But the devil is in the detail. With data collection and algorithms running on auto pilot, they can end up averaging a huge quantity of measurements down to a few numbers. Errors and unwanted information can be buried within a mass of records. Big data sets, with millions of records make it easier to force a correlation that isn’t necessarily accurate. This paper will briefly reflect on the positive reactions and outcomes to our paper at the 2017 AITPM National Conference, how we have taken the advice further and what that means to our future research.
The paper will present some of the latest trends that are embracing the input of more data into transport planning and management including a specific session at the recent 2017 International Conference on Survey Methods in Transport.
It will then address a number of specific transport planning measures, such as the development of trip tables and the need to eliminate real errors. It will also highlight how pressure to drive more numbers into the matched tables also creates distortions.
Finally, the we will give some reflections on how transport professionals must apply wisdom, born from experience, to ensure that ideas, patterns and solutions reflect the real world.
CONCURRENT SESSION 7: Modelled Integration
The main objective of this study is to examine a wide spectrum of primary determinants that explain both spatial and temporal variations in public transportation use in Perth, thus offering a comprehensive and cogent analysis of public transport use in Perth metropolitan suburbs. It constructs a novel and comprehensive database of public transit in Perth by integrating multiple databases on land use characteristics, geo-spatial information, socioeconomic conditions, urban form factors, public transport service provisions at the suburb level along with the revealed preferences
derived from smart cards.
The Darlington Project is a $620 million road infrastructure upgrade which forms part of the long-term vision of a North-South Corridor for Adelaide. Darlington contains the busiest road segment (Main South Road~73,000vpd) and busiest at-grade intersection (Main South Road and Sturt Road ~98,000vpd) in South Australia. The project is grade separating five existing signalised intersections across a 2km stretch of Man South Road with eight road bridges being built. The project runs through a precinct which contains the second largest hospital in SA, the second largest University in SA as well as the largest shopping centre in SA close by. The scheme has to contend with extremely complex movements with two roads feeding and receiving traffic at either end, causing traffic movements to be incredibly complex and dynamic and multiple peaks within peak periods due to differing traffic movements. The scheme currently being built is not the scheme from tender time. Subsequent to the tender, traffic volumes were updated which required the scheme to be updated to accommodate the significant increase in traffic forecast to move through or travelling to the precinct in the peaks. This paper will discuss the development of the scheme, the principles ultimately adopted, accounting for the complexity of movements in a safe and sensible manner and the challenges in trying to meet the project objectives; all the while knowing that construction in the incredibly constrained area would play a significant role in ensuring the traffic does not suffer catastrophic delays during construction. During the development of the scheme, issues surrounding the consistency between assessment packages (Sidra/HCM/Aimsim) caused considerable time and effort to be required to reconcile differences and provide the best possible advice..
The principle objective of this paper is to assist in the critical appraisal of unvalidated microsimulation models and the questions to consider when procuring such a model. Very often pedestrian microsimulation models are built for infrastructure that only exist in the future, this usually leads to skipping the validation phase as there is no existing comparable element upon which to base a validation model. In this case, how can the robustness and reliability of the model results be determined and when presented with results, how should they be interpreted? Many of the issues arise from a poorly specified role for the model. This paper will demonstrate how a loosely worded specification can result in modelling results which comply with the specification but would provide a poor customer service or even a potentially unsafe environment.
CONCURRENT SESSION 11: Operational Modelling
The presentation gives modellers, project managers, promotors and agencies an overview of the Operational Modelling Guidelines and Audit (OMeGA) document developed by Main Roads Western Australia, explaining the roles and responsibilities of the individuals involved in the model submission and the benefits of following the audit process for Main Roads’ regulatory approvals.
CONCURRENT SESSION 15: A Question of Mode
This research and think piece compares mode choice and topic public transport assignment globally. A better detailed understanding of the differences may lead to improvements for all locations. As an example, in the US, the typical approach encompasses an elaborate mode choice model and a relatively simple public transport assignment. In Europe and Australia often the converse is true.Background, findings, conclusion: Many US transport models use nested logit mode choice, with multiple access submodes (walk, drive, park-and-ride, kiss-and-ride) and multiple public transport submodes (local bus, express bus, BRT, light rail, heavy rail, commuter rail). This produces a complex model with many possible choices. Given the relatively low level of public transport usage in most of the US, it is nearly impossible to estimate such a model from home interview survey data. Typically, many of the coefficient values are asserted and the bias coefficients adjusted to match observed subtotals. In some nests, the bias coefficients are fairly large, limiting the model’s sensitivity. However, this approach allows the model to match the observed trips closely by loading onto the public transport network separately by submode.
Newcastle Light Rail has adopted a catenary free project solution whereby power supply occurs at stations, whilst the vehicle is stationary and passengers are boarding and disembarking.As power supply is linked to power usage, and power usage is linked to light rail operations, the determination of power supply needs has become intrinsically linked to the operations of the light rail vehicles. Furthermore, operations of light rail vehicles can and will vary depending on traffic delay which can create situations in which multiple vehicles are simultaneously charging at stations.
To evaluate this dynamic, a sophisticated traffic model was used to appraise the light rail operations with respect to infrastructure provisions, traffic signal operations, light rail priority and catenary free power supply. The design process required a co-ordinated process between traffic modelling, light rail vehicle manufacturers, power supply designers and contractor.
This paper explores alternate methods of modelling future land use, with a focus on forecasting land use which will best inform project evaluation and policy. It highlights the benefits of using a methodology which combines theoretical frameworks with forward looking supply data and policy direction.
CONCURRENT SESSION 19: Efficient Networks
Traditionally, in the UK and globally, road infrastructure projects are justified through economic benefit delivered in the AM and PM peak hours. Average hour models are constructed to build business cases, with home-based work commuting trips generally accounting for the majority of the benefit. So what happens when the network peak does not coincide with the traditional commuter peaks? Average hour models no longer deliver the level of detail we require to provide a justification for investment and there is a requirement for something more comprehensive to improve the assurance we can provide to decision makers. Could the use of disaggregate journey time data hold the key?
Modelling managed motorways interventions is not a new development, however their implementations mainly exist in microsimulation models. Whilst these types of models are very capable of modelling managed motorways interventions, their detailed operational nature makes them less suitable for longer term planning. Another limitation exists in their restrictions in modelling large scale networks due to prohibitively long simulation run times and high costs to setup.These limitations are very much reduced in macroscopic dynamic assignment models whilst still providing appropriate levels of detail for planning purposes. Their strengths are in scalability, the relatively low computational requirements and their deterministic outputs. The outputs of these models are well suited for assessment and appraisal of managed motorways interventions. The StreamLine framework in OmniTRANS which uses the MaDAM propagation technique is one such model. This modelling system incorporates many managed motorways interventions like ramp metering, variable message signs (VMS), lane use management, and variable speed limits (VSL). Coordinated ramp metering was recently added following the HERO (HEuristic Ramp metering cOordination) feedback control strategy in combination with the ALINEA ramp metering algorithm.
CONCURRENT SESSION 23: On Space Cost and Time
Transport models become more complex and untamed every year. A full run of a model can take days to complete and can provide details of every modelled trip to several degrees of precision. The purpose of this paper is to highlight the areas of modelling we seldom think about: the definition of a trip, the meaning of the trip matrix, the definition of the day we are modelling, and many more. These are the foundation of our models and can demolish the purpose of our forecasts if they are not carefully considered and explained.View Presentation
Geographical Information Systems (GIS) have long been used to aid transport and land use planning, but are you aware that the analytical and visualisation capabilities of GIS have come a long way in recent years? A GIS is much more than just a digital map. It’s a system that can integrate datasets from a variety of sources, manipulate them and provide timely outputs to support transport and land use decision makers. Does this ability sound familiar to you as a transport modeller? Are GIS software and personnel now in a position to impinge upon the analysis typically performed by transport modellers in transport modelling software? Should we be scared? Or can our GIS counterparts be our allies?