robotaxis vs autonomus minibuses
Automated transport of people can be used in two ways, and the difference decides what a city gets. In public transport, it addresses a real shortage. It makes it possible to run public transport vehiles despided the growing shortage of human drivers and automus minibuses can serve thined routes at ten in the evening, or feeder service in a suburb where a scheduled line has never been economic. Autonomus minibuses extend the network instead of replacing it, and per passenger it uses a fraction of the street space. Robotaxis do the opposite. Both will happen. Which one dominates depends on rules that do not yet exist – and that is the opportunity rather than the problem. For once the rules can be written before the market forms instead of after it.
what robotaxis bring
None of what follows is misbehaviour. Every effect below is the rational response to the incentives a city currently sets: circulation is free, stopping is scarce, the kerb is unpriced, the charger is unallocated and nobody is obliged to hand over data. Change the incentive and the effect changes with it – which is why this is a list of design decisions rather than a list of complaints.
on people
growing up in a chauffeured city
independence is built, not granted
A child who walks to school, cycles to a friend, takes the bus to a club is not just getting from one place to another. They are learning to read a street, judge a gap in traffic, ask a stranger for directions, notice that they have gone the wrong way and work out how to get back. None of that is taught. It is acquired by doing it, in small increments, from an age at which the increments are small enough to be safe.
A child driven everywhere arrives at the same places and learns none of it. The destination is reached and the capacity is not built – and the capacity, not the journey, was the thing that mattered.
this has already happened once, without anyone deciding it
The decline is documented, and the figures are worth stating because they are larger than people expect. A long-running comparison of England and Germany recorded the share of primary schoolchildren allowed to travel home from school on their own:
| allowed to travel home from school alone | 1971 | 1990 | 2010 |
| England | 86 % | 35 % | 25 % |
| Germany | — | 91 % | 76 % |
The same study found that of English seven and eight year olds, 80 % were allowed to go to school unsupervised in 1971 and 9 % in 1990. Two things follow. Germany is in a much better position than England and is moving in the same direction – 91 % to 76 % in twenty years is not a stable situation. And nobody chose this. There was no decision, no vote, no policy. It is the accumulated result of a street environment that became more dangerous and more unpleasant for a child on foot, one road widening at a time, until parents individually concluded that the reasonable thing was to drive.
Shaw, B., Watson, B., Frauendienst, B., Redecker, A., Jones, T., with Hillman, M. (2013): Children’s Independent Mobility – a comparative study in England and Germany (1971–2010). Policy Studies Institute, London.
the service removes the pressure, not the cause
A robotaxi fleet does not cause this. Car dependence caused it, over fifty years, and the damage was done long before automation existed.
What an always-available, cheap chauffeur service does is worse in a specific way: it removes the last pressure to reverse it. At present a parent driving a child to school pays for it – in time, in the morning, in the second car, in the twenty minutes sitting in the queue of other parents doing the same thing outside the same gate. That cost is uncomfortable, and discomfort of exactly that kind is what produces the demand for a safe route, a school street, a crossing that works. It is the reason these things ever get built. Take the friction away and the demand goes with it. The journey becomes easy, the cause remains untouched, and the condition stops being experienced as a problem at all. Nothing is more durable than an arrangement that has stopped being annoying.
and it feeds itself
Fewer children out alone means fewer children in the street. A street with no children in it stops being a place where children are expected, so it feels less safe to send one out – and the parent who would have allowed it concludes, correctly for their own child, that it is unusual and therefore risky. Fewer children again.
This is the same self-reinforcing loop as speed and sprawl, running on a different variable. Each individual decision is sensible. The accumulated result is one nobody chose and no one person can undo.
what that adds up to
a less capable generation, then a less capable society
Independent mobility is not only about transport. The child who has never navigated a street alone has also never had an unsupervised encounter with an adult who is not family, never had to decide anything without someone to ask, never been mildly lost, and never discovered that they could handle it. These are the ordinary conditions under which people learn that they can act. An adult who has not had them is not incapable. They are, on average, more reliant on being carried – by a service, by an institution, by someone else’s arrangements – and less practiced at the plain competence of dealing with the world as it is. One generation of this is a change in the population. It shows up, eventually, in whether people organise anything themselves, help a neighbor, come forward in an emergency, or expect that someone else will have handled it.
and a more isolated one
A life conducted door to door – private space, private vehicle, private space – contains no unplanned encounters. Nobody is met who was not arranged in advance.
Shared space produces contact as a by-product of ordinary movement: the same faces at the same stop, the neighbour recognised from the pavement, the conversation nobody intended to have. That is how a street, a square, a library or a bus works as social infrastructure, and it is cheap precisely because nobody has to organise it. Remove ordinary movement from shared space and the by-product stops as well – not through anyone’s decision, and with nothing to put in its place.
Loneliness is already rising, most sharply among the old and the young. A transport arrangement in which the young are chauffeured from an early age and the old have no reason to be on the street does not cause that, and it removes one of the few remaining mechanisms that works against it.
This is why societal integrity is one of the protected assets in this model rather than a soft consideration alongside the real ones. A population that has had little practice at being in public space together is harder to inform, harder to organise, harder to evacuate and slower to help itself. That is an operational safety matter, and it is produced by planning decisions that are never assessed against it.
the objection, taken seriously
There is a real counter-argument and it should not be waved away. For a child in a suburb with nothing in walking distance, an affordable automated ride could expand their range considerably. A twelve-year-old who currently reaches a club only when a parent is free could reach it whenever they like. That is a genuine gain, and for children with disabilities it can be a large one. Two things about it. It is independence from a parent’s schedule, which is worth something, and not independence in public space, which is the thing that builds the capacity described above – being driven by a machine develops nothing that being driven by a parent did not. And it is an argument for making the route work, not for buying the ride: the same money spent on a safe path and a crossing gives the child both kinds of independence and leaves the street better for everyone else who uses it.
That is the accessibility question in its sharpest form. Before buying mobility, examine whether the distance can be made walkable – because the ride solves the journey and the path solves the childhood.
the missing driver
A driver is also a witness. Someone who notices a passenger who should not be travelling alone, helps with a pram, intervenes when something happens at the back of the vehicle, or is simply present at a stop at one in the morning. An automated service replaces that with cameras and a remote operator, and for some passengers that is an improvement – there are people who feel safer without a stranger at the wheel. For others it is a loss, and it falls on exactly the groups least able to absorb it. Whether the substitution works is a design question with an answer, and it is one the operator should have to demonstrate rather than assert.
blocked public charging points
In San Diego, where Waymo began commercial operation in September 2026, electric car owners report waiting between half an hour and more than an hour at public fast chargers because robotaxis are using the stalls. Similar reports come from Las Vegas, Nashville and St. Louis. The charging operators state that their public stations are open to every driver, robotaxis included, and that no special arrangement exists. They are right, and that is the point. Public charging infrastructure is built, and usually subsidised, so that people who cannot charge at home can run an electric car at all – people in dense housing without a private driveway, the households with the least space and often the least money. A fleet charging at those stalls breaks no rule, because when the chargers were planned the only users imagined were residents. The result is a transfer from the people the infrastructure was built for to a company that can build its own, and it happens without anyone taking a decision. That is the ordinary pattern of public space appearing in a new place: whoever arrives first, at a price unrelated to what is taken, until someone writes a rule.
the people next to the depot
A fleet has to stand sometimes somewhere, and the somewhere is a yard in a residential district, because that is where the land near the demand is. The noise, the lighting and the movement at night fall on whoever lives besides it.
In Santa Monica, residents beside two charging lots say the vehicles’ reversing alarms run more or less continuously. People report not being able to sleep or open a window, have petitioned the council, and in some cases have blocked the entrances with traffic cones or with their own bodies. The operator sought police intervention and a restraining order. City enforcement measured the noise and found it did not breach the local limits.
Both findings are true at once. A noise limit is derived for the sources that existed when it was written, and it is averaged over a period; a reversing alarm repeating at short intervals through the night is a different kind of exposure, and can sit well below an average limit while making a flat unusable. The limit is not being cheated. It is measuring something else.
the ratchet
prices that rise when the alternatives thin out
Dynamic pricing peaks when riders have the fewest options: rain, night, after an event, during a strike. The moments at which a public service is most needed are the moments at which a private one charges most, and that is not a flaw in dynamic pricing – it is what dynamic pricing is for.
By the time the effect on the public network is visible in the figures, residents have arranged their lives around the new service. People have given up a car, chosen a flat, taken a job on the assumption that it will be there. At that point withdrawing or heavily conditioning it is no longer a transport decision but a political one, taken against the people who depend on it. The window in which a city can set terms freely is the window before anyone needs the service – which is also the window in which it is hardest to argue that terms are necessary.
more traffic on the streets
more vehicles for the same journeys
A robotaxi drives to its passenger, drives the passenger, and then drives on to the next one or circles while it waits. Experience with ride-hailing in US cities shows more vehicle kilometres than the journeys themselves require, a large share of them empty. Empty running is not a flaw in that model. Where circulating costs nothing and stopping is scarce, circulating is the cheaper of the two, and an operator that chose otherwise would be paying for a courtesy nobody asked it for. The behaviour follows from the price of standing still, and so does the remedy.
induced treffic
A service priced below its cost for a decade does not only take journeys from other modes. It creates journeys that were not made before – the trip someone would have walked, combined with another, or not taken at all. Every transport improvement that lowers the cost of a journey has done this; there is no reason this one would be the exception. For the operator that is growth. For the street it is more traffic, and it arrives first in the places that are already busiest, because that is where the service is worth running.
blocking of the kerb
Pick-up and drop-off happen at the kerb – and where no kerb space is designated, they happen in the cycle lane, at the bus stop, across the dropped kerb at a crossing. Each of those is somebody else’s space, and the three somebody-elses are a cyclist, a bus and a wheelchair.
This is the ordinary competition for street space with a new and well-capitalised participant. What makes it different from a delivery van is volume and persistence: a fleet is at the kerb all day, every day, by design.
the approach for the emergency services
More vehicles in the carriageway and more occupied kerb length both lengthen the approach to an incident. That much is simply arithmetic.
The second part is not. How a fleet behaves towards a vehicle on blue lights, whether it keeps clear of an incident scene, and whether it still does so when the network is down are questions type approval does not answer. The potential here runs the other way too – a fleet that always reacts the same way could clear a path far more reliably than people do – and that is examined in its own section below. Either way it is decided by the conditions of access, not by the technology.
on the public transport
Robotaxis go where it pays: the center, and the busy hours. That is the same traffic a public network relies on to cross-subsidise the thin routes, the late evening and the outskirts – the services that will never cover their own cost and are the reason a public network exists at all. So the loss is not proportional. A fleet taking ten per cent of the passengers takes them from the ten per cent of services that generate the surplus, and what gets cut in response is the night bus, not the trunk line. The outskirts and the night stay uncovered, and become harder to fund than they were before the service arrived.
missing traffic data
Trip data shows where and when people actually travel – the most useful planning information a city could hold about its own territory. If the operator keeps it, the city plans its network without it, and negotiates without knowing what it is negotiating about. The asymmetry compounds. Each year the operator knows more about the city’s mobility than the city does, and each year the city’s own counter-proposals rest on older data.
dependence on a handful of platforms
If a few global operators come to dominate, a city depends on their prices, their software and their servers for a service its residents rely on. That dependence is not hypothetical in either direction: a platform can raise prices, change its coverage, or leave a market, and a city has no substitute ready because it stopped building one.