What Happened to Delivery Drones?
On an episode of 60 Minutes in 2013, Jeff Bezos introduced the world to the concept of drone delivery that was under development at Amazon. It was predicted in this interview that within five years, drones would populate cities and guarantee 30 minute deliveries on otherwise multi-day Amazon fulfillments. This look into the future sent the world abuzz and spawned the creation of start-ups such as Zipline and Skydrop who began partnering with FedEx and UPS to implement drone pipelines. It felt like the famous “last mile” efficiency problem that plagued all delivery services finally had a modern solution.
Present day however, we are coming out of a two year pandemic in which demand for delivery over brick and mortar purchases skyrocketed, yet the delivery process is still mostly manual. Our packages are delivered via truck and our food orders by gig workers with large cubes on their back. Amazon Air meanwhile, closed its UK office in 2018, laying off all employees and providing no further updates to this day. Is delivery drone rollout simply taking longer than Bezos thought, or was Amazon’s vision impossible from the start?
The benefits of drone deliveries are pretty straight-forward:
- Reduced shipping cost
- Reduced ground traffic
- Reduced emissions
- Reduced delivery time
Lower cost and emissions are especially exciting since the last mile of deliveries account for ~40% of total shipping cost and high greenhouse-gas emissions. Successful deployment of drones could achieve “anything, anywhere, anytime” delivery with these added sustainability bonuses. Underneath Amazon’s original drone concept however, lies an iceberg of logistical and tech barriers.
The logistical problems start with the Federal Aviation Administration (FAA): a part of the US government that decides who has jurisdiction over what airspace. Most conversations about drone deliveries center around residential applications, where drones deliver parcels to individuals like an automatic mailman. Due to limited batteries, these use cases are restricted to high density areas such as cities and their surrounding suburbs. A perfect city to consider for drone deliveries would be Phoenix, Arizona due to its stable climate and sprawling suburbs that cause wasteful deliveries otherwise. Phoenix, like every other large city in America, has multiple airports and an airbase who all claim exclusive airspace via the FAA (see Phoenix Airspace Map below). Flying drones in these spaces is out of the question, which leaves only 40% of Phoenix’s airspace available for delivery.
Even within the available airspace, the next question for drone delivery becomes where should the drone land? The ideal scenario for a delivery is a single-family household with a spacious lawn for the drone to release the package. Since companies like Amazon and Skydrop are considering their drones for city usage, they’ll need to address cramped, high density apartment buildings and skyscrapers as delivery destinations. Will all packages be left at the same doorstep, or will drone companies have to collaborate with landowners to create drop-off zones?
Beyond solving where to fly drones and where they should land, the problem that stumps all engineers is how to safely land a drone consistently. When ordinary people were asked in a 2020 logistics survey what makes them most anxious about drone delivery, the number one answer (21%) was damage to packages/recipients. Ensuring no harm to a drone package given the numerous possible drop-off locations is a huge design issue. Allowing even a single delivery to cause harm would destroy national trust in drones and trigger lawsuits, so companies cannot afford subpar drone intelligence.
Creating an environmentally aware drone means capturing sensor data as input to computer vision techniques such as image recognition, route-planning, and most importantly risk assessment. An intelligent drone needs to decide in real-time whether or not their delivery zone is safe for landing. There are unlimited edge cases that make evaluating this risk difficult. For example, if a drone spots a person within a radius of their landing zone that it deems unsafe, how does it know whether that person is outside, or watching the drone from behind a window in their house? Or what if right before the drone reaches the ground, a dog jumps out and charges the drone? Some of these hypotheticals may sound ridiculous, but companies like Amazon looking for massive drone deployment need to consider all scenarios.
Due to these design hurdles and many others, COVID 19, which could have been the beginning of Bezos’ drone revolution instead saw the explosion of delivery apps into a $150 billion industry. Over the past three years, companies like UberEats, Doordash and Grubhub have enlisted an army of uninsured gig workers to complete last mile deliveries with incredible speed. People have shown that they’re willing to pay more for this speed, which should’ve been a good omen for drone companies, but Amazon and others have still barely given a proof of concept for aerial delivery. Even so, the low-tech solutions by delivery apps satisfy two of people’s greatest concerns with drone delivery: stolen packages and job replacement. It has become increasingly clear recently however, that the gig economy on which delivery apps operate is not built to last.
For one, gig deliveries require bloated service fees for the consumer and restaurant using their services. That fee unfortunately doesn’t trickle down to the deliverer themselves who are often underpaid. Even for the app companies themselves, profits are becoming increasingly thin. UberEats and some of its competitors are coming full circle by investing in automated deliveries to boost their business viability.
Automation for companies like Postmates (now part of Uber) are coming in the form of land vehicles. Their “Serve Robot” model is about the size of a rolling suitcase and shares the sidewalk with pedestrians to deliver within a three mile radius. Keeping deliveries planted on the ground avoids conflicts with the FAA, lowers safety concerns for dropping packages and increases user experience since most bots include a padlock screen for recipients. Although a bot moving at pedestrian speed isn’t a threat, a unique challenge for these vehicles is terrain scanning. What happens when there is no sidewalk to travel, or a poorly paved sidewalk knocks the bot on its side? For this reason, trials of the Serve bot and one of its competitors called Starship have found great success on college campuses, where routes are consistent, and sidewalks plentiful. Bot-resistant cities like San Francisco have taken notice of this progress and permitted Postmates’ sidewalk robotics.
With the forward march of land delivery vehicles, it’s worth looking back on Jeff Bezos’ original vision of 30 minute aerial deliveries and considering whether it was a pipe dream. Maybe instead, his vision had inflated expectations that stemmed from a hype cycle. The Gartner Hype Cycle (see below) is a model of technological adoption that models how a new idea is treated as it enters the public sphere. It starts with a technological trigger that sparks media coverage and eventually leads to inflated public expectations. From there, the technology fails to meet those impossibly high expectations and the cycle reaches a “trough of disillusionment”. Crawling out of that trough involves finding the realistic use cases of a product and finally proving the idea’s worth. Recent drone successes have shown that the industry may be digging itself out of disillusionment.
The drone company Zipline who ran into the same computer vision blockades as Amazon have found an effective niche for aerial deliveries: medical supplies. West African countries such as Rwanda run community medical clinics that accept walk-in patients seeking vaccinations, antibiotics and other supplements. These clinics are faced with a design challenge to keep ample supply of vaccines, which spoil easily and require refrigeration. Since the clinics often can’t afford to stockpile vaccines, a practical solution has been on-demand drone deliveries. Volunteers keeping inventory at the clinics can order medical supplies exactly when they need it, and a Zipline drone travels from a distribution center to drop the payload into a designated landing zone. Giving drones a static landing space removes a layer of intelligence needed for drones to replace Amazon deliveries. Although the damaging of packages is still a risk, Zipline and Matternet have offered solutions like parachutes attached to a parcel or slowly lowering the package with a cable.
Investment companies have begun taking notice of the humanitarian drone uses, and could support future deployments. Maybe now that research is beginning to hone in on practical use cases for drones, we’ll see a more measured version of Jeff Bezos’ vision that ignited the world.
