Innovate Gur Kimchi will bet on drones It is the only form of freight transport that can guarantee the meeting between speed and economy. A dialogue with Gur Kimchi, the visionary engineer who invented Amazon Prime AirPioneer and visionary. These are the two characteristics, beyond the academic qualifications and the rich professional curriculum, that best describe Gur Kimchi. He brought his ideas to Microsoft, to Waze, to 5 startups he launched. The only limit? The sky. And here that too becomes a challenge. Joined Amazon in 2012, he gave birth to the Prime Air project, for the use of drones in deliveries. Gur is a founding member of the FAA Drone Advisory Committee, and worked in collaboration with the FAA, NASA, SESAR, and ICAO on the development of the Federated Airspace Management Architecture, enabling the safe integration of drones into the airspace around the world.What was a drone when you started studying the project and what is it today?When we started working on what ended up being Prime Air, the initial objective wasn’t to build drones. What we did was to “work-backwards” from a defect we recognized in the transportation system. We forecasted that as customers continue to order more of their purchases online, and expect these orders to arrive faster, the stress that these more-and-faster deliveries place on our transportation networks will become extreme and unacceptable. To solve this defect, we decided to search for the fastest, safest, most-economical, and environmentally-responsible delivery method for the last mile, while simultaneously reducing the future stress on the existing networks. We ran many experiments and simulations on a wide range of transportation modalities, recognizing that no single solution will satisfy all customer needs. The simulations affirmed something we knew intuitively: the current road-based transportation system suffers from two fundamental challenges that will make it harder to scale while consistently delivering faster. The first challenge was that the dimensionality of the road network is somewhere between 1 (imagine moving along a 1D line) and 2 (which will be moving using the shortest-path between any two points on a 2D plane). The road network dimensionality is obviously larger than 1, as we do manage to get to our destination, and the surface of the earth is for most day to day practical purposes a flat 2D plane. But consider that we cannot traverse the road network along the shortest-path line. One has to travel from A to B while respecting the road’s various constraints, so it is clear that the road network dimensionality is somewhat less than 2. There is a direct relationship between the scalability of a given transportation network and its dimensionality: the larger the dimensionality, the more intrinsically scalable a given system can be due to shorter paths and fewer conflicts between concurrent users. As an example, by this metric bicycles are more scalable than cars, simply because fewer constraints are placed on where bicycles can go. Bicycles are still limited to a network dimensionality of <2 as they cannot safely cross highways or go through buildings, for example, but they can be used to build somewhat more “scalable” networks compared to cars which are limited to driving only on the roads. The ultimate expression of this dimensionality metric is the Air, which is 4-dimensional: users can share the airspace as long as they don’t conflict in X, Y, Z and Time – change any of these parameters enough and a conflict can be avoided, and direct paths are not only possible but generally preferable. This means that any transportation system that operates in the air has the potential to be much more scalable than anything on the ground. The second challenge was that any system that operates on the ground has to interoperate with other users on the ground, which would be us – as drivers, pedestrians, bicyclists etc. The problem is easy to see when we consider Self-Driving cars – imagine we built a perfectly clean and protected road network dedicated only to self-driving cars – the technical and societal challenges would be much easier! Our conclusion was that isolated networks can offer performance, scalability, simplicity, and safety advantages that integrated networks simply cannot, and that the air is a lot more scalable than the ground. Consider then that the airspace above us – generally over 200 feet above most buildings, and up to 400 feet, leaving a 100 feet buffer with most of general aviation except of course close to airports on the takeoff and landing corridors, is essentially empty. It is an isolated slice of air that is currently underutilized. This was the point when we started to seriously consider drones. I came into the project with enough historical aerospace experience to have a huge amount of respect as to how hard it is to build a safe, certified, commercial aerospace system, and early on I tried as hard to find a solution that did not require us to build airplanes, but we realised that the advantages were so huge that we just had to try. “Any transportation system that operates in the air has the potential to be much more scalable than anything on the ground”What distances does it make sense to cover today with drones for the transport of goods? I generally think about drones being useful in locations that are underserved: too far from commerce centers, or local commerce centers lack sufficient inventory to get you what you need. Rural locations are an obvious place to start, and are a focus of many players in the nascent industry. Suburban communities are also commonly underserved due to traffic – the store may be nearby when there is no traffic, or often when there is no traffic – such as late at night the nearest store may be closed…. There’s another advantage to delivery by drones: it has the potential to save us from having to context-switch. Even if I could drive my car (or better, ride my bike) to the nearest store and get what I need in 30 minutes, I still have to stop whatever it is I’m doing. It is preferable to have the things we need come to us, vs. us having to go to them – assuming of course that the cost and the environmental side-effects are not negative – and with pure-electric drones, it’s all positive. This is analogous to the transformation we have experienced over the last 15 years with digital contents such as books, music and videos – we no longer have to go to the music, video or book stores: the switch to digital streaming services allows us to “pull” media through smartphones or smart speakers wherever we happen to be. In other words, “things” come to us – instead of us having to go to them. Drone delivery is simply the same idea extended to physical items. Getting to your range question – it turns out that flying even short distances – say 10-15 km – is quite useful due to the fact that drones travel directly to their destinations. For a given route, the driving distance can be as much as twice as far as the flying distance. There is also a useful upper limit to range – once a drone can fly, say, 50-75km – essentially any customer can be reached from a few central locations. While I suspect we will first see drone deliveries in rural and suburban locations, i’m not saying we will never see drone deliveries in Cities – it may happen at some point especially for time-sensitive or emergency goods. I expect that these Urban Drones will drop and pick up packages using automated lockers deployed on the roof of highrise buildings – but we have to consider that Urban customers can probably get to many local stores almost as quickly as a drone delivery, so the advantages are not as material. But I dont think package delivery will be the first Urban drone application – I predict we will see many other incredibly useful drone applications deployed much sooner – from life-saving emergency response supporting fire and medical personnel, to safety monitoring of air and water quality, to inspection of safety-critical infrastructure such as bridges, roads, or dams. I also think these applications will end using many different types of drones built for fit their specific purpose – and I expect that these mission-specific systems will look very different from what we think of as drones today. “In a first phase, the urban drones will deliver and collect the parcels using distributed automated lockers on the roofs of skyscrapers”What are the main advantages you see in this form of transport?I like to think about the advantages of drone delivery as a combination of capabilities: the ability to enable rapid delivery, often in less time that it will take someone to go to a physical store, without having to increase the costs the customers will have to pay, without having to compromise on the environmental benefits, without having to further overload the existing ground-transportation network, at great scale – etc. Here’s another way to think about this combination: in today’s transportation networks, we often have to choose between – for example – speed and cost. If we need to deliver something quickly, we know that the price will go up, and the inverse is of course also true – if we strive to reduce the cost of a delivery, we expect the delivery to take longer – in other words, it is speed OR cost. This OR is the main issue solved using drone delivery – we can replace these ORs with ANDs – we can deliver fast AND economically AND environmentally-cleanly AND scalably – I consider this switch from OR to AND to be incredibly impactful. “In today’s transportation networks, we often have to choose between speed and cost”The use of drones in urban centers it may envisage intermodal solutions, where the aircraft will be entrusted ‘the last mile’ of the deliveryWhat kind of infrastructure do you need? One way I think about this question is: will there be many drone operators each serving a specific business-application and customer population with a high degree of overlap, or do we expect there to be fewer. While I dont expect drone delivery to be a winner-take-all business, I do expect – mainly due to the regulatory complexity related to aircraft, manufacturing, autonomy and operational certifications – that we will end up with a fewer and larger operators who are probably more global vs. local. Will there be 3 or 10 operators per major region (say, north america or europe)? Is not a question I think we can answer right now – but I’m pretty sure we will not have 1 or 2 – the market is simply too large – but I don’t expect will we end up with, say, 10,000 small local operators. With that in mind, I believe we can make some predictions about what infrastructure will be needed to support these networks. I predict we will end up with 2-3 operating centers per operator, per metro area, staged in the industrial neighborhoods. We could call these logistical centers airports but I expect they will end up looking quite different from today’s airports with their runways and terminals. I expect these sites to hold as much inventory and internal sortation capabilities as possible – at the limit we can imagine any customer order being fulfilled by any available drone, which will be flying to any customer location directly with that one package. So what will this infrastructure end up looking like? I’ve seen photos of hive-like structures, vertiports, and many other possible designs. I predict that each operator will design a unique, fit-for-their-purpose site that is optimized end to end, and that these optimization may end up being per site. Some operators may build permanent sites, while others may go with modular or mobile sites. In other words, the differences – the variability – between operators and sites may end up being much bigger than the similarities.by Simona Vecchies (from “DomusAir” n. 1) “Each operator will design a unique, fit-for-their-purpose site that is optimized end to end, and that these optimization may end up being per site” back to top