The promise of drone delivery has hovered on the horizon for more than a decade, but only now is it beginning to look commercially viable.
Advances in regulation, autonomy, and cost-efficiency are transforming drones from experimental novelties into serious contenders for last-mile logistics.
At the centre of this shift is the broader field of uncrewed technology – a sector that encompasses everything from aerial drones to maritime systems and ground robots, and which is increasingly shaping how goods, data, and services move across modern economies.
Few people are better placed to explain this transformation than James McDanolds, program chair at the Sonoran Desert Institute’s School of Uncrewed Technology.
With a background that bridges aerospace, instruction, and practical deployment of uncrewed systems, McDanolds brings both technical expertise and a sharp awareness of the regulatory and economic pressures that determine whether a technology takes off – literally and figuratively.
In this conversation, we discuss why Uber’s new drone delivery partnership could succeed where its earlier Uber Elevate project struggled, how the Federal Aviation Administration’s proposed Part 108 rules may accelerate beyond-visual-line-of-sight (BVLOS) operations, and the role of AI in scaling autonomous systems.
We also explore safety, reliability, and the hard realities of power consumption in an era of rapidly expanding AI data centres.
What emerges is a clear picture of an industry on the cusp of change: moving from one-pilot-one-drone experiments to multi-aircraft networks that may redefine not just e-commerce, but the very infrastructure of the skies.
Q&A with James McDanolds, program chair at Sonoran Desert Institute’s School of Uncrewed Technology
Robotics and Automation News: Uber’s renewed push – Uber Elevate faced challenges in gaining traction, but you’ve said this new partnership with Flytrex could succeed. What’s different this time?
James McDanolds: Primarily the proposed part 108 ruleset and not just BVLOS but Multi-UAS primarily. I detail the whys a bit more in the response to the next question.
R&AN: The FAA’s forthcoming BVLOS (Part 108) framework is seen as a turning point for drone delivery. What are the key changes, and how might they accelerate large-scale deployment?
JM: One of the larger differences now compared to previous endeavors is the scale in which the ease of operational permissions are given. In the past BVLOS operations were still approved by the FAA on a case-by-case basis.
Now with the notice of proposed rulemaking for part 108 that is very much tailored with Drone delivery style, at scale operations, it makes it more possible to see a potential return on a service like this at scale in a field that mainly consists of small sized orders with aircraft that can handle it.
In the past BVLOS operations, even under waiver and if not already combined with a Multi-UAS waiver required a pilot per aircraft on a one-to-one ratio.
This includes the salary of the pilot, the cost of the aircraft, and the cost of the equipment to operate it, which the cost of the salary and equipment increases the cost of the delivery in order to get a return.
Now under part 108 as it has been proposed a singular operator, translates to a singular salary, can monitor multiple aircraft. Which means one operator to multiple drones which lowers the cost of running the operation.
So now it’s one salary plus one set of ground control equipment to operate 20 aircraft instead of 20 pilots’ salary with 20 sets of ground control equipment to control 20 drones.
Potentially reducing the cost of the operation by 2/3rds and making it feasible to incorporate without having to get approvals for each new area of operation that they want to conduct drone deliveries in.
R&AN: You’ve highlighted AI autonomy as the real enabler of scale. How far along are drone systems today in handling complex, real-world environments without human pilots?
JM: Drones, even without AI and just using the autopilot firmware’s out there that currently exist such as Ardupilot and PX4 already have a level of intelligence that allows for autonomous flights to be conducted in different profiles such as mapping, spraying, drone delivery, and more without the level of AI that we have seen come to the field in just the last two years.
But many of those flight plans have to be planned by the pilots flying them for different applications. Same with monitoring the aircraft during its flight as a flight without a human in the loop is not allowed, even in the currently proposed part 108 rules.
Maybe in the future, but for now, even with a high level of autonomy with our current AI, the national airspace system is complex and there is much it would need to account for and build up enough data to validate the exclusion of a trained pilot or “Flight coordinator” being involved in the operations as a backup.
What current AI can do to help advance further autonomy is the faster generation of flight missions (using the current waypoint programming in firmware’s I mentioned earlier) so that pilots don’t have to spend hours if not days building the autonomous missions.
Which makes conducting flights to millions of different addresses more feasible if the drone has not flown to them before for a delivery.
R&AN: From avoiding power lines to parachuting packages, what technologies or protocols are proving most effective in making delivery drones safe and trustworthy?
JM: Collision avoidance is the biggest key, not just visual based, but when considering more drones for delivery being flown on the day to day at scale, they will have to also consider other aircraft flying in a newly cluttered airspace, even at sub 400ft above ground level altitudes.
For example, Zipline and Amazon are flying in the same city. How do their systems talk to each other to be able to monitor each other’s drones or flight paths and altitudes to ensure they don’t cross each other to make sure there is not a mid-air between drones from two different service providers operating in the same area?
We could have designated flight altitudes for each service provider, but that only works is the number of service providers in the area are limited in number due to the available vertical airspace altitude range and the limit that each may want to set to allow for altitude deviations if needed in case of emergencies.
R&AN: Despite progress, issues like power grid demand, privacy, and airspace congestion remain. Which of these do you see as the biggest obstacle to mainstream drone delivery, and how can the industry overcome it?
JM: Besides what I mentioned in my response your earlier question, I do see the increased power demand becoming an issue over time, especially with the large increase on the power grid from increased and rapid expansion of AI data centers have had on the market recently which in some cases have seemed to trickle down to local homeowners increasing their costs per KWH.
Small drones’ power could be offset by solar as they are small enough where a solar panel could be used for battery charging as long as some of the other supporting equipment such as power banks are there, but that only works for a certain level of operation and scale for anyone area.
For example, two solar panels could be good enough for a single aircraft at a hamburger spot delivering hamburgers themselves with a drone, but an army of panels would be needed for an amazon warehouse launching hundreds of flights per day.
