The Last-Mile Delivery Problem Nobody's Solving
Key Takeaways
- Last-mile autonomous vehicles could cut costs by 40-60%, but lower delivery costs will trigger higher delivery volumes.
- The real challenge is managing the demand for limited street space.
- Last-mile autonomous vehicles come in many forms (sidewalk robots, drones, vans, cargo pods), but we should not try to predict which will win; instead, focus on the outcomes we want, regardless of the technology.
- Without proper demand management, autonomous deliveries risk degrading public spaces through increased congestion, noise, and obstacles for pedestrians and cyclists.
- The solution is smart pricing: charge autonomous vehicles for using roads, footpaths, cycle lanes, curb and air space.
- Digital coordination of curb space can transform it from a chaotic free-for-all into a managed resource where vehicles book slots, pay appropriately, and move on efficiently.
- Revenue from space pricing should fund infrastructure improvements, wider footpaths, protected cycle lanes, and road diets, creating a virtuous cycle.
- Cities should not rebuild infrastructure around delivery robots; autonomous vehicles must adapt to existing crossings, signals, and street designs.
- Cities face a choice: let autonomous delivery evolve unchecked and sacrifice livability, or use pricing to align delivery efficiency with the cities we want to live in.
What Next?
Next week, I will pull together all six parts of this AV series to help create a vision of how we should get the most out of AVs.
Introduction
Over the past few weeks, I have discussed AVs from a range of different perspectives. Last week, I looked at long-distance freight. This week, it is the turn of last-mile freight deliveries.
Last-mile autonomous vehicles are already here. They are delivering groceries in Milton Keynes, transporting medical supplies in San Francisco, and, for those of a romantic bent, even helping people propose to their girlfriends.
But while many celebrate the technology, few anticipate the second-order effects and ask: what happens when delivery costs drop by half?
The answer is simple and worrying: we will get far more deliveries.
Long-distance freight is relatively straightforward: replace human-driven trucks with driverless trucks. Last-mile is different. We are experimenting with sidewalk robots, delivery drones, autonomous vans, cargo pods on bike lanes, and mobile parcel lockers. Each has different infrastructure requirements. Some use roads, others use footpaths or cycle lanes, and drones need no street space at all.
This diversity reflects how complicated last-mile delivery actually is: different goods, different urgencies, different destinations, different constraints.
So which solution will win?
We do not know. And that creates a problem when you are trying to set out a long-term vision.
In this post, I will focus on what we can probably predict: the benefits, the downsides, and the framework we need to get the most value from autonomous deliveries while avoiding the worst outcomes.
I am deliberately avoiding short and medium-term issues like safety regulations, technical development, and labour displacement. Those matter, but they are transitional challenges. I am focused on the longer-term vision, the world after autonomous delivery becomes widespread and cheap.
That world is coming fast, and we have shown no sign of being prepared for it.
Last-Mile Delivery AV benefits
Advocates for last-mile delivery AVs paint an appealing picture. The benefits they promise are significant:
Dramatically lower costs
Removing human drivers could cut operating costs by 40-60% for last-mile deliveries. Given that last-mile delivery typically accounts for half the total cost of getting goods to customers, this represents a massive efficiency gain.
Solving labour shortages
Recruiting delivery drivers is increasingly difficult. Autonomous vehicles sidestep this constraint.
Round-the-clock operations
Deliveries can happen whenever it makes operational sense.
Faster deliveries
Without human shift limits and with better route coordination, delivery times shrink.
Smarter routing
Autonomous vehicles can optimise routes in real-time far better than human drivers. They can also use infrastructure delivery vans cannot, cycle lanes, footpaths, and even airspace in the case of drones to bypass congestion entirely.
Better use of curb space
Instead of delivery vans circling blocks hunting for parking spots, autonomous vehicles can coordinate their curb access digitally. Book a spot, use it, move on. No more cruising, no more blocking traffic.
Cleaner deliveries
Electric autonomous vehicles accelerate the shift away from polluting delivery vans.
More reliable deliveries
Today, about 5% of deliveries fail, wrong address, nobody home, access problems. Better coordination and communication could dramatically improve this success rate.
Lower prices for consumers
When delivery costs drop, those savings should flow through to cheaper goods and delivery fees. That means real cost-of-living improvements for households.
These benefits are compelling. The problem is they only tell half the story.
What this list focuses on is the efficiency of individual deliveries. What it ignores is what happens at a system level when you make deliveries radically cheaper and more convenient.
Because when something gets 50% cheaper, people use a lot more of it. And that is where the problems begin.
Potential Last-Mile Delivery AV Downsides
The benefits sound great in isolation. But zoom out to the system level, and a different picture emerges.
The induced demand problem
Cheaper deliveries mean more deliveries. We have already seen this pattern with e-commerce: delivery traffic has surged as online shopping has become easier and cheaper. Now imagine cutting delivery costs in half again.
If those deliveries use roads, we get more congestion. If they use footpaths and cycle lanes, those spaces become increasingly crowded. Either way, our infrastructure gets more stressed, not less.
The battle for curb space intensifies
Delivery vans already compete fiercely for curb space with parked cars, loading zones, and cycle parking. More deliveries could make this worse.
Yes, autonomous vehicles might reduce overall demand for parking as passenger AVs drop people off rather than park. And yes, smaller delivery robots might not need parking at all.
But those gains could be overwhelmed by sheer volume. More deliveries mean more vehicles stopping at curbs, even if each stop is more efficient.
Public space degradation
Streets are not just transport corridors. They are public spaces where people walk, talk, play, and live.
Flood those spaces with delivery robots and drones, and you degrade the quality of urban life. More obstacles for pedestrians, particularly those with disabilities. Less space that feels genuinely public and calm.
The real issue nobody wants to face
All these downsides share a common thread: they are about what happens when you do not manage demand for limited space.
The benefits list assumes each delivery gets more efficient. The downsides list recognises that more efficient deliveries create far more deliveries, which can overwhelm the efficiency gains.
This is the gap between microeconomics and systems thinking. Optimising individual trips is easy. Optimising the entire system when everyone is optimising their individual trips is hard.
And right now, few people are talking about this problem.
Infrastructure Constraints
If we are going to incorporate last-mile AVs into our cities, we need to think hard about infrastructure. Not just roads, but footpaths, cycle lanes, curbs, and crossings.
Put enough delivery robots on footpaths and cycle lanes, and those spaces become less useful for their primary purpose, moving people.
Managing demand, not just supply
There are two ways to solve crowded infrastructure: increase supply or manage demand.
We will need some of both, but demand management has to come first. Without it, we simply induce more traffic to fill whatever capacity we create.
This is where road user charging becomes essential. I have argued throughout this series that charging for road use will be critical for managing autonomous vehicles. The same logic applies to footpaths and cycle lanes.
Charge robots for using these spaces. Charge more during peak pedestrian or cycling times. This does three things:
First, it prevents the system from unduly favouring footpaths and cycle lanes over roads, which would distort routing decisions and overload pedestrian infrastructure.
Second, it manages demand by making delivery companies optimise their use of limited space. Deliver when paths are quieter. Consolidate trips. Choose routes that minimise impact.
Third, it generates revenue to expand and improve the infrastructure. Use the proceeds to widen footpaths, create better cycle lanes, and fund road diets that reclaim space from cars.
Curb space coordination
The curb is already contested territory. Cars park. Delivery trucks stop. Cyclists need protected lanes. Cafes want outdoor seating.
Autonomous vehicles, both passenger and delivery, offer a path forward: digital coordination.
Instead of first-come-first-served competition with vehicles circling for spots, AVs can book curb space in advance. They pay for the time they need, use it efficiently, and move on.
This requires dynamic pricing. Curb space in busy areas at peak times costs more. Quiet streets at off-peak times cost less. The price signal guides vehicles to use curb space where and when it is least scarce.
Done right, this transforms the curb from a source of chaos and cruising into a managed resource. Everyone gets better outcomes than today's free-for-all.
The limits of infrastructure adaptation
Some advocate for redesigning infrastructure to accommodate autonomous vehicles, adapting pedestrian crossings so robots can trigger them, creating dedicated lanes, installing sensors and communication systems.
This is backwards. We should not rebuild our cities around the needs of delivery robots.
Autonomous vehicles need to adapt to existing infrastructure, not the other way around. If a robot cannot handle a standard pedestrian crossing, that is a robot problem, not an infrastructure problem.
Yes, some targeted improvements make sense: better curb management, perhaps dedicated loading zones in high-traffic areas. But the default should be that AVs solve their own navigation challenges.
Cities have limited budgets and competing priorities. Spending heavily to accommodate delivery robots while public transport needs expansion is a choice we should scrutinise very carefully.
What about drones?
Drones present a unique case; they use airspace rather than roads, footpaths, or cycle lanes. This might seem to exempt them from infrastructure pricing. It should not. Airspace in urban areas is a limited resource just like street space, but the constraint is different: noise and visual amenity. A few drones, if quiet enough, might be barely noticeable. Hundreds overhead constantly could be a problem. Pricing drone flights manages this amenity impact while preventing the system from artificially favouring aerial delivery over ground-based alternatives.
Without airspace pricing, delivery companies would route everything through the air simply because it is free, even when ground delivery would be less disruptive overall. The principle remains the same: charge for the use of shared resources, whether that resource is pavement, road surface, or the sky.
Informing a vision
The next few years will be dominated by testing, safety regulations, and pilot programs. That is necessary work, but it does not constitute a vision.
A vision needs to look past the transition period to the steady state, the world where autonomous delivery is mature, widespread, and cheap. What do we want that world to look like?
Start with outcomes, not technologies
A trap we could fall into is trying to predict which technologies will win. Will it be sidewalk robots? Drones? Autonomous vans? Some combination?
We do not know. And we do not need to know.
What we need is clarity about the outcomes we want:
- Lower delivery costs that translate to genuine cost-of-living improvements
- Livable cities with streets that feel pleasant and safe for people
- Sustainable transport with minimal environmental impact
- Reliable deliveries that work for businesses and consumers
- Efficient use of limited urban space
- Equitable access to goods and services
Notice what is missing from that list: any mention of specific vehicle types or delivery mechanisms.
The vision should be technology-agnostic. We care about results, not methods.
Build rules that deliver outcomes
Once we know what we want, we can design rules that push toward those outcomes without micromanaging how delivery happens.
This means abandoning the temptation to pick winners or mandate specific approaches. No rules requiring certain vehicle types. No bans on technologies. No subsidies for favoured solutions.
Instead, create rules that reward good outcomes and penalise bad ones.
Conclusion
Last-mile autonomous vehicles promise real benefits: lower costs, faster deliveries, and round-the-clock service. But these advantages come with a catch that many people are overlooking.
Cheaper deliveries mean more deliveries. And more deliveries mean more pressure on our already limited street space, whether that's roads clogged with delivery vans, footpaths crowded with robots, or bike lanes shared with autonomous pods.
The technology will improve, costs will fall, and adoption will accelerate. The real challenge is ensuring our cities don't become paralysed by their own efficiency gains.
We don't need to predict which type of delivery robot will win. We don't need to redesign every crossing and curb for the perfect autonomous future. What we need is the right framework.
That framework centres on three principles: charge for the use of limited space, coordinate access digitally, and invest the proceeds in better infrastructure for everyone. Smart pricing for roads, footpaths, cycle lanes, and curb and air space is about ensuring our public spaces work for people.
The decisions we make in the next few years will determine whether autonomous deliveries enhance our cities or degrade them. Fail to put pricing in place, and we'll have efficient logistics but unlivable streets. Get it right, and we can fund the wider footpaths, better bike lanes, and improved public spaces that make cities thrive.
Next week, I will complete this series on Autonomous Vehicles to help create a vision of how we should get the most out of AVs and improve our lives.