Exploring the Potential of Journey PlanningPaul Everson | April 30, 2015
ChoiceRail mixed mode planner development report
Journey planning solutions have unquestionably come a long way in the past 20 years – yet they remain fundamentally limited by the assumption that users want to plan a journey either with a private car, or with public transport solutions. In reality, most journeys can be fulfilled using a combination of the private car and public transport. Shouldn’t our public transport journey planners take this into account?
That was the underlying theory that gave rise to ChoiceRail, a development co-funded by the Technology Strategy Board, and delivered as a collaboration between Trapeze Group, Cotares Ltd. and the University of the West of England.
With ChoiceRail now complete, this article reviews the project, reports on its findings, and ultimately aims to answer the question, what is the true potential of mixed mode journey planning?
Where we are today?
Journey planning solutions have existed for some twenty years. Yet, when planning a journey today, travellers are still faced with the dilemma of identifying which options are available, often from multiple sources, relying on any background knowledge and their own judgement to decide what might best suit their personal preferences. Journey planners on the market today typically offer either private car or public transport solutions. A few claim to offer both, but almost without exception this means a comparison between public and private transport.
I may be over-simplifying the usage here, but I expect many modern car drivers tend to enter destination details – either into a Smartphone or in-vehicle ‘Sat Nav’ – as a matter of course when starting their car journey, and trust the planner to get them to their destination. Such devices often now include real-time traffic feeds to alert drivers of upcoming congestion and offer alternative routes without any requirement to stop and think.
By contrast, public transport journey planners are still primarily used as tools for pre-trip planning – and as a result they rarely track progress during the journey. They have included real-time data for some years, but are only now beginning to offer genuine real-time planning.
Social Media is becoming widespread as a means to broadcast disruption information, but rarely tells the traveller how to ‘fix’ their journey.
The mixed mode journey planning premise
At Trapeze, we wondered whether there is a market for a mixed mode planner – and in particular, one that integrates road with public transport. If such a market exists, what would an integrated solution offer? Which journey types are best achieved using a combination of car and public transport? We assume the car is always quicker than public transport. And cheaper too.
We had already identified that there were a wide range of data, algorithm and emotive issues to answer, and that integration is a complex problem to solve. We therefore focussed on one particular use case and tried to answer the following question: If I am doing a longer distance journey, what are the optimum mixed mode (car – rail) solutions? We realised this came down to one key decision: what is the best interchange station to use?
Current public transport planners can’t answer this question – for a given origin and destination they are algorithmically restricted (for reasons of reducing the response time to the query) to search for solutions that use stops close to the origin and also the destination. The consequence is that they fail to evaluate interchanges further afield.
A well-known example of the problem we were trying to solve is a journey from Northampton to Newcastle. Most planners will offer solutions that include public transport routes via London – with resulting journey times an hour longer than driving. Intuitively, we believed users would expect to see a mixed mode solution that includes a drive to Peterborough. This scenario is widespread – inferior solutions because the planner forces you on to the public transport network close to your origin without considering the entire journey.
When journey planners go wrong
The ChoiceRail project
During the ChoiceRail project, we worked with two key partners – Cotares, a Cambridge based SME with a novel patented real-time road routing algorithm, and the Centre for Transport and Society at the University of the West of England (UWE) who are established leaders in the understanding of the links between lifestyles and personal travel behaviour.
Technically, the solution seemed quite simple:
- Use Cotares’ road algorithm to find candidate interchange stations
- Create and populate a data model for car parking and modal interchanges
- Use the Trapeze public transport algorithm to find solutions from the candidate interchanges to the destination
- Build a set of results that removes duplicates and promotes diversity of results
The Cotares algorithm is able to find car journey times from a given origin to all 2000+ stations in the UK in less than half a second. Journey times are based on real-time road data giving an extra dimension of credibility.
Early on in the project it became clear that this approach generated far too many candidate interchange stations – so the work focussed on identifying key ones. This then lead to further questions about how to decide which are the best interchange stations – what is the car parking capacity, and is there availability at certain times of the day? What facilities exist at these stations? What frequency are the connecting rail services?
It was evident that these questions are key to the decisions made by ‘real’ people when considering their travel options. We realised that some of the data didn’t exist that would enable us to answer these questions – or was commercially unrealistic to include. We resorted to using some pre-determined default values in lieu of any better data. As a result of the work undertaken, we were able to determine a set of journey options which could be presented to the user.
In parallel to the development of the software demonstrator, UWE undertook some research which identified four notional journey types. These reflected the options that we expected the ChoiceRail algorithm to generate:
- Drive only
- Local station rail – journeys ostensibly by rail where the point of access to the rail network is close to the traveller’s origin – these are the options that existing journey planners provide,
- Split modes – journeys which are distinctly part by car and part by train and notably where the share between modes is 50/50, and
- Park and Ride – journeys ostensibly by car but where the traveller stops at a station short of their destination in order to continue the last part by train, bus or tram
At this point in the project, it wasn’t yet clear which of these journey types the ChoiceRail algorithm would offer. We recognised the risk that when your public transport algorithm is allowed to offer car, that “drive only” would always come out as the timeliest option. This wouldn’t have been an ideal outcome – not least because our clients are providers of public transport, or are Local Authorities encouraging modal shift towards public transport.
UWE also looked at the types of travellers who might be interested in ChoiceRail as a solution. An initial set of focus groups highlighted five different traveller behaviours in relation to how road and rail are used in combination in practice:
- Determined car driving – travellers wedded to driving for long distance journeys
- Local station train loving – travellers who adopt ‘no brainer’ train use for long distance journeys but who use a car to access ‘their’ station
- Evolutionary exploring – travellers who develop their mental map of options for road-rail completion through experience and insight from significant others
- Habitualised innovation – travellers who have come to use ‘non-local’ rail access for specific long distance journeys which no longer seems ‘peculiar’ to them
- Unsuspectingly disrupted habit – travellers who have preconceived assumptions about mode options but who may be ‘nudged’ into considering new options
More questions were raised: Did we need to tune the algorithm or user interface to target one (or more) of these behaviours? Did we need to ensure the ‘expected’ journeys were present alongside the alternatives offered by ChoiceRail?
In order to answer these questions, a demonstrable user interface was required that surfaced the ChoiceRail algorithm. Trapeze developed a website using Responsive Web Design principles and HTML5, which would offer a high quality user experience regardless of the device type used. This decision was based on analysis of the usage of our existing client journey planning solutions, which showed roughly equal numbers of users accessing via mobile devices and traditional desktop devices.
We were encouraged to see that ChoiceRail offered the expected answer to the Northampton to Newcastle query. We were also encouraged by positive feedback from users in that it – more often than not – detected the interchange that users had established through their own choices. We were especially encouraged that some mixed mode solutions were no greater in duration than the car only alternative; indeed we found many examples where the overall journey time was quicker using a mixed mode solution.
A logical journey plan
Using the prototype, UWE gathered feedback from a second round of focus groups. This testing exposed three key insights.
- There are use-case scenarios where participants could envisage the benefit of an inter-modal journey planner
- ChoiceRail can be used as a way of ‘exploring’ travel options, and as a way of uncovering new ways of making a journey
- Thirdly, this work gave some indication of how likely people were to actually use ChoiceRail. Issues such as complexity of solutions and trust in results were amongst factors highlighted in this context
On this last point, there were many factors which were identified, but would be hard to model in our algorithm:
- If the interchange station is in an unfamiliar location, some users would be reluctant to select that option
- The confidence that there will be a car parking space (in addition to cost and capacity)
- The facilities at the interchange station
- The connection to the train leg. What frequency are the trains if I miss the intended connection? Will I get a seat on the train? How much will it cost?
- Will I be able to get back to my car (is there a return public transport option)? Will the car park still be open when I get back?
In summary, ChoiceRail allowed us to prototype a mixed mode planner. We were reassured that the algorithm returned the results expected by users. We were equally excited to discover that the algorithm found the interchanges that reflect real users’ choices.
We hope the results of our algorithm will encourage people out of their cars and on to public transport for a significant part of their journey. The user interface we developed will overcome the issue of the user having to juggle several road and public transport planners on the screen – which would be near impossible on a Smartphone screen.
Perhaps most importantly, we realised that some parameters that are most important to travellers are not yet modelled, nor the data available. Twenty years on from our first attempts, we are still occupied in the pursuit of the perfect planner – one that reflects both the objective and subjective decisions made by real people.