Riding the Carrot train
“To achieve great things, two things are needed: a plan, and not quite enough time.”
24 hours is a good interval to breed and slaughter ideas.
A decade is a good interval for permanent change: We spend nine years agreeing to change and how, and then the last year (or day) actually changing.
What is this document about?
The carrot framework looked at how to get to a sustainable transport system without introducing exotic new technology, and it claimed that today’s system is not sustainable.
We proposed to merge existing line-oriented transport with adaptive transport, supported by partially-existing information systems.
Definition: Words and phrases like “point-to-point”, “real-time”, “dynamic”, “adaptive” and so on are quite the mouthful. They can’t be completely avoided, but to simplify I have made up a word just for this document: “Flexride”.
A flexride is a trip from origin to destination using an optimised sequence of [point-to-point, real-time, dynamic, adaptive, etc, etc] rides.
The carrot framework is using soon available information to shift from today’s transit system towards one built on flexrides.
Wasn’t the talk enough?
It was a start, but given “not quite enough time”, it was very unfinished. This document is intended as an intermediary step, the annotated version. You will find four types of annotations, highlighting hand-waving, references, diversions and hooks. (The section numbers follow the talk.)
Our prime technique to reach the deadline was hand-waving, “let’s just assume this will work”.
With more time at disposal we can show it is realistic that it would, maybe even how.
📚 References & expansions
For those of you that don’t think this document is long enough, I might add further references and relevant documents at a later date.
Ideas and thoughts pruned during the Hackathon, that might be worthwhile to pursue. We had many ideas, this is far from a complete list.
Some of these are relevant to this talk, others may be relevant in other contexts, and yet others may be buried with a minimum of fuss.
What we can use this for, potential projects. If you don’t have time for anything else, skip to these.
The transit from 2020 to 2030
The annotated slide show.
👐 1.1 Climate change and emission targets
Climate change is real, and will increasingly affect our future behaviour on transport, agriculture, energy supply etc. But we have not progressed very far on our path from awareness to targets to policy. At some point we will change not just because we want to, but because we must.
Today we may pick A over B because A is more sustainable. Tomorrow B might not be an option. In our time frame we need to entice people to do the right thing, we cannot enforce (11.3).
Our talk assumed our awareness would lead to policy change (e.g. in Trafikverket). This cannot be taken for granted in practice.
📚 1.2 Emission target impacts
Even so, the targets are supposed to be reached somehow. We can document a disconnect between public emissions targets and current policy (but not in this document).
📚 1.3 Sustainable Development Goals
We did not link up to the 2030 Agenda for Sustainable Development. Even without the added focus on carbon emissions this is a worthwhile project for other reason. We could document other ways this could add value.
👐 3.1 Why NOT public transport?
There are studies for why people prefer not to use public transport, even though it would be better for the society.
This document will continue to hand-wave, perhaps linking to said studies at a later stage.
👐 3.2 Time and the rest
The talk focused on one important factor: Time use. Today public transport comes in most cases, and almost all cases in Stockholm county, at a time premium. If you live in the countryside, or if you live in Södertälje, this premium is huge. If you don’t have a car, you will spend much more time in traffic.
If we had more than five minutes at disposal, we could expand on all the other factors that make people pick one mode of transport over another.
👐 3.3 Driver psychology and motivation
This project is partially motivated by the climate benefit of drivers switching to public transport. A switch from walking, bicycling or staying home would not have any effect.
The assumption is that if the alternative is fast and convenient enough people would switch from driving otherwise empty cars. This assumption does not generally hold.
The same applies to other approaches like car sharing. In principle car sharing, where up to 5 people used a car instead of just the driver, would be the cheapest way to cut emissions. No new investments in public transport or infrastructure necessary. In practice car sharing has not been a great success. Commercial sharing is taxi by another name.
Most prospects for future self-driving cars are essentially car sharing too, only without the driver. They may not end up differently from today’s schemes.
🎣 5.1: SL orders lines years ahead by public procurement
This is a crux of the talk. We cannot have an adaptive/dynamic system if we only procure predetermined lines, stops and frequencies in advance.
We need to work out a system to target transport needs at any given moment, much like taxis do for personal transport and logistics companies do for packages.
In addition to procuring pre-determined services (“who will charge the least to provide a bus service along this route with these stops, and a frequency of 15 minutes”), we propose to procure transport resources (e.g. a number of buses and drivers) to be used for “just-in-time transport”.
👐Without such a system we claim we can’t shift from current unsustainable modes of transport without an unrealistic investment in public and personal transport.
👐 5.2 Predictability vs adaptability
A trunk line time table system is very predictable, if you know the lines and time table, and the time table can be kept without delays, or the frequency is high enough that the time table doesn’t matter.
An adaptive system is not predictable for those not enrolled in it, but can be more predictable for those who are.
👐 5.3 Poor data quality
The planners can only know which lines passengers are using when, but not where they originally came from and wanted to go, nor if there was a route they would have preferred to take over the one they actually took.
In actual fact planners would know even less, at least in Stockholm (7.2).
The response time to the data also suffers, as the data gathering – analysis –procurement cycle takes years.
👐 6.1 Line-oriented vs real-time point-to-point
The goal of this project: More adaptive use of public transit, particularly for those who today rather drive cars.
Transport agencies order or set up (train, tram, bus…) lines for public transport based on past experience. These lines and frequencies are published. Systems on top of that, e.g. Google Maps or Resrobot, use this to find the shortest path for the traveller.
Point-to-point recognises that mobility is not actually a sequence of transports (“take pendeltåg 40, wait 5 minutes, then bus 758”), but a “desire path” from here (e.g. home) to there (e.g. work), possibly with intermediate stops.
Real-time recognises that even with a given schedule, real-time issues (traffic, accidents, delays) can affect the optimal transport route.
A journey based on this information is called a flexride.
📚 6.2 The Internet is a template as well as a tool
Early telecommunication was line oriented. To send information from one node to another a line was set up for the purpose until the transmission was done.
The most revolutionary IT change that led to the Internet was packet switching. Instead of setting up connections, information was marshalled into addressed packages using and adapting to the available communication channels.
Public transport has not yet made this shift, while logistics companies largely have. Their job is also to deliver packets through existing transport channels, routing and re-routing based on current condition.
👐 6.3 An adaptive system should work well with line-oriented transport
Flexrides have to coexist, use and improve existing line-oriented transport. Rail, metro and light rail will largely follow a schedule, and the rails. The same goes for trunk-line buses, effectively rail-on-wheels.
There might be some leeway to adapt the schedule to real-time traffic (e.g. more trains during a football march), but the backbone traffic is likely to be regular and predetermined.
This will somewhat constrain flexrides. If the next train leaves in 20 minutes, this would be a given even though the adaptive transport could fill it in 10.
👐 6.4 Time-optimising and route-optimising are similar
Within constraints, we know how to find the quickest route. That is not quite the same as finding “the fastest for the mostest for the cheapest”. Optimisations can be found though. Just a small matter of programming.
👐 7.1 Open data sources
Public data is the largest source of information, and to a large extent publicly available.
Commercial (above “private”) data are useful both as primary sources (above Scania) and as value-added services (e.g. Google Maps and OpenStreetMap).
There are issues with user-owned data, such as privacy and what would convince or incentivize them to offer them up. The mobile phone will have almost perfect travel record, so Google (Apple) knows, the rest of us are more or less in the dark.
Legal changes during this decade could provide another source of data. There is a proposal to make vehicle taxes based more on actual travel distance and depending on which roads (e.g. cheaper on countryside roads than city streets). Such a framework would in effect register all car journeys in the country, at least temporarily. If there was a way to access these data without breaking anonymity or privacy laws, we would have information on all journeys affecting (1.2), possibly as input to (8.1).
👐 7.2 Traffic data
We assume real-time and aggregated travel data is available.
Use of SL cards tell where and when traveller get on, but not where they got off or the route they took to get there. Stations (but not bus stop) may have the number of exits, but the exits can’t be linked to the entries.
This can be augmented by traffic surveys on board or after, but surveys are costly and incomplete, and not directly mapped to transport data.
🎣 7.3 Search data
Searches are “desire paths” from where you are to where you want to go. Traffic data tells how the buses and trains are used, search data can make sense of these by knowing from and to where the travellers are going.
The time table data are available to e.g. Google Maps. One possible licencing model would be to request aggregate data (i.e. from their searches) in return, for better future transport planning.
Either way there could be data quality issues, in principle the search data could be gamed.
A search could be turned into a flexride. There are already limited opportunities to do so from Resrobot (and other transit providers). This could be further enhanced (see 11.1).
🔜 7.4 The NMT of transport data
The 2G public telecommunications network was based on GSM. That in turn was based on experiences from NMT, Nordic Mobile Telephone. Like happened with NMT, Nordic cooperation could be a stepping stone for European/global standards for exchange of transport data.
Whatever the outcome, the transport providers should be in contact with like-minded providers in other locales.
🎣 8.1 Learned patterns
Machine learning could “fill in the gaps” of our incomplete transport knowledge.
Actual transports could be re-run in simulations to see how far they diverged from optimal transport.
Either way these could be useful tools for planning both line transport and flexrides.
👐8.2 Open APIs
The data produced could and should be available for outsiders through public and documented APIs. The licences could be used to improve not only (), but possibly also data quality (7.3).
The system could become a marketplace for service providers outside SL (e.g. car sharing), buyers of flexrides (travellers, but also possibly companies for their employees, customers, visitors), and value-added-services.
It might be useful not only to provide open APIs, but also open source. City transport is a shared problem, and the solutions could also be shared (7.4).
🎣 8.3 Planning and visualisation
It is probably quite some time away for systems in (8.1) to replace transport planners, but these data could certainly help them.
This visualisation could be directly supported, or added by third-party through the API (8.2).
Visualisations could help planners see not only where the system is underperforming, and has potential for further optimising (alternatively better city planning), but also how well it is competing. For Carrot particularly against driving (3.1), to fulfil the environmental goal (1.2).
Visualisation could help the traveller as well, both changing to flexriding and finding optimal ways to travel.
Roads not taken
Several ideas and approaches were pruned in the process. Here are most of these. Section 11 are elaborations, 12 and 13 are fairly independent projects (for Södertälje and the region respectively).
🎣 11.1 End-to-end flexride
Searches (as in 7.3) can be used to buy tickets, but today this hands-off ends the transaction. Instead a flexride should be considered completed when either the traveller reaches her goal, reaches a changed goal, or cancels the rest of the trip. In other words, the flexride ends when she is at her “there”.
It will adapt to delays and changes in the transport system, but also delays and changes for the traveller. The bus† will know at any given moment how many flexriders want to enter the bus and where, and if they are likely to reach that bus. The flexrider will know which sequence of rides to take, and delays and proposed changes as applicable, waiting time and where/if she has to hurry (or pick a less stressful route).
† for “bus” replace with whatever mode of transportation is applicable (the digital shadow of that “bus” to be exact).
🎣 11.2 Quality of service
Not just the fastest/cheapest way to reach the destination, but also require a certain quality of service, corresponding to the requirements in (3.1).
Office workers might want to work during the commute, and might want to e.g. dynamically reserve seats for the flextrip, as could people with mobility issues.
They might need a guaranteed time of arrival: “I have to be there at 9:00 or (I will cancel this ride | I will be very unhappy | I will have to make a phone call | I want compensation (13.2))”
👐 11.3 The stick
The original framework had both the carrot and the stick, with the emphasis on the former.
We dropped the stick for reasons: The carrot is more effective psychologically, it is more pleasing and profitable to do the transition in (1.1) by making the alternative better rather than current transport worse, and finally the stick is unlikely to make the deadline. In our case the 24 hours/5 minutes, but also the 10 years to 2030.
It is ultimately up to the politicians, but we can’t expect any major legislation restricting car use before 2030. The most ambitious target I know of world-wide is Norways goal (1.1) all new cars should be zero-emission by 2025. Even if this target is reached, the existing vehicle fleet will mostly be non-zero in 2025 (but the proportion will fall towards zero by 2030). And while “zero-emission” cars emit less CO2, they still emit indirectly through power use and production cost.
The stick approach may have negative social impact for those on the wrong end of it.
Cities can have growing restricted “car-free” zones, and built-in advantages and subsidies (like free parking) may slowly disappear. Car sharing might be supported (3.3), or single-driver trips restricted.
2030 is unlikely to be dramatically different to 2020, though the direction is clear: Unnecessary high-impact travel will be increasingly strongly discouraged.
A softer stick would be a shift in transit priorities (13.3).
👐 12.1 Södertälje the gateway to Stockholm County
Assuming Stockholm got an efficient transit and transport system, there would be no need for travellers going to Stockholm county to go further than Södertälje before switching to this system.
👐 12.2 The transport modes divide the city
But Södertälje as gateway suffers from being separated from itself by these modes of transport (sea, road, rail). The motorways, railroads, the Södertälje Canal and the harbour, divides Södertälje into isolated neighbourhoods.
Apart from overcoming roads, rails, and water as physical barriers we need to overcome mental barriers. Someone living in one district has little inclination or motivation for visiting another. Somebody from Rosenlund is unlikely to visit Ronna, and vice versa. Cultural exchange between Mariekäll and Saltskog or Saltskog and Hovsjö is also unlikely.
There should be something in each district to attract visitors from the rest of the city, and there should be city streets and footpaths that entice people to explore.
👐 12.3 Getting around in Södertälje is difficult and slow
A common complaint at the Science Week was that getting around in Södertälje is as hard and slow as getting to Södertälje in the first place.
(I can confirm that. I live fairly central in Södertälje. Still, the time it takes to walk to the bus stop, wait for the bus, take the bus, and wait for the train can be as long as the train ride.)
This is what Carrot and flexrides are intended to solve. But it would be worthwhile to look at why travel in Södertälje is slow, even between places that physically are fairly close. Both for the travellers’ convenience and for strengthening the social fabric of the city.
👐 12.4 The last mile problem
As mentioned above Södertälje getting from major transport hubs to home, or simply getting around, can be a challenge, even with relatively small distances (1–3 km).
Buses may take travellers “to the door”, but there are practical and legal reasons why this may not be a practicable idea.
1–3 km is a feasible distance to travel by bicycle, electric wheelchair, hoverboards, and other means of personal transportation, but usually too long and time-consuming to comfortably get there by walking. (Bicycles and e-bikes can of course travel much further than 1–3 km, but not the issue here.)
Personal mobility is more “social” than most other modes of transport. In most cases transport separates us from our neighbours and environment, we are not here when our goal is to be there.
🔜 12.5 Goods and long-distance transport
Traditionally local transit, long distrance travel and goods transport use separate systems. They need not do so. For out-of-towners these considerations may affect their choice of transport (e.g. plane, train or car).
Södertälje can be a practical conduit for all of this (12.1).
🔜 12.6 The Stockholm Gate
On the idea of Södertälje as a gateway (12.1), and as a way to rejoin Östertälje, I have earlier proposed a BRT (13.3) joining destinations in Stockholm with Östertäljes enklaver, Moraberg & Stockholmsporten (in Swedish).
🎣 13.1 Payment model: minimal cost card
With a heterogeneous group of service providers and flexrides, the payment model could end up more complex than today.
The best answer to this would be that the transport system continuous (re)calculates transit fees to the user’s maximum advantage. If a day card is cheaper than single (flex)trips, the day card fee is deducted. Likewise for the other options over the transport system, particularly when non-SL trips are included.
🎣 13.2 We here for you (Förseningsersättning)
A variant of which is delay compensations. SL paid out 29 MSEK in compensations 2019, not including the cost of managing the service. Today this system is very cumbersome for the customers, and probably for SL.
Compensations could be calculated automatically, and the most co-advantageous solution found. By lowering the barrier SL would be liable to pay compensations for far more delays, but each delay could become much less costly, as the system could offer alternative transport instead of compensations up to 1100 SEK.
🎣 13.3 Priority lane (gräddfil)
Proposed target for 2030: Where public transport does not have right-of-way in Stockholm region it shall have priority.
In practice this would mean that any road with more than one lane shall have a priority lane. Trafikverket is sceptical to priority lanes, as there is much more private than public transport. Reducing the number of lanes by one will to some extent increase traffic on the remaining lanes, and thus travel time.
This is a line-oriented approach (6.2). A packet-oriented approach would maximise the use of the priority lane by allocating surplus capacity. If a stretch of road can handle 200 extra cars, these slots can be allocated by e.g. real-time auction, queuing, or a real-time priority system.