.png)
The UK drone sector has spent the last five years celebrating the wrong milestones. Every extended-VLOS approval, every observer-chained corridor, every Temporary Danger Area gets announced as a breakthrough.
But an approval is not an operation, and an operation that requires a chain of people standing in fields is not an industry.
Visual mitigation (BVLOS with visual observers, extended VLOS, daisy-chained handovers) has become the default path to "BVLOS" in the UK. It is the wrong default. Not because it is unsafe when done properly by good operators, but because it fails on all three of the tests that actually matter: it does not close the safety argument it claims to close, it inverts the cost curve of the technology, and it cannot scale.
The CAA itself has been clear about the destination: to scale successfully, UAS "will need to be able to integrate safely and routinely into UK airspace, without need for 'special provisions' or 'segregated airspace'" (CAP3182, Future of Flight: BVLOS Roadmap). Visual mitigation is a special provision made of people.
Here are four reasons why it should not become the default model for BVLOS.
.jpeg)
1. Observers make the economics worse as operations grow
One of the fundamental advantages of uncrewed aviation is that the cost of an additional flight can fall significantly as operations scale. Visual mitigation works in the opposite direction: the longer and further the aircraft operates, the more people are needed to support it.
The numbers are stark. McKinsey's analysis of observer-constrained short-range delivery put direct operating cost at roughly $13.50 per delivery, of which up to 95% was labour. That fell to $1.50–$2.00 when a single operator could manage 20 aircraft (McKinsey analysis, Jan 2023).
Their conclusion was that until BVLOS becomes normal practice, "the present business is not economically viable due to the need of a visual observer and the short flight distances this creates."
The UK offshore case shows the problem particularly clearly. To achieve BVLOS with visual mitigation over water, operators put an observer in a boat, at a typical cost of £2,000 to £3,000 per day (Drone Regulations and Net Risk, p. 21).
That single cost can exceed the entire capital amortisation of the aircraft for the sortie.
The same document records that BVLOS powerline inspection was 34% cheaper than VLOS, while autonomous operation can cut inspection costs by up to 70%.
Visual mitigation, in other words, is not a cheap bridge to scalable BVLOS.
2. It puts more people into the risk environment
This is the argument I care most about, because it is often missing from the regulatory discussion.
The purpose of uncrewed inspection and logistics is, in large part, to take people out of hazardous environments. Drones "do not eliminate risk. They relocate it away from people" (Applying Net Risk: The Safety Case for UK Drone Regulations).
The effect can be substantial. If an inspection takes one pilot one day instead of a four-person crew five days, human exposure falls by 95%. The dominant residual hazard can then be something much more ordinary: "driving to site is usually one of the riskier activities on a drone RAMS" (Drone Regulations and Net Risk, p. 16).
Ground crews also bring their own risks and controls.
UK highways guidance requires a dedicated crew member positioned near live traffic to monitor it and retrieve a fallen aircraft before it becomes a road hazard, with high-visibility PPE (UC Drones Policy Guidance). National Highways GG 954 also treats drone activity near roads as a potential distraction to road users (Standards for Highways, GG 954).
This matters because many of the environments where BVLOS creates the greatest value are precisely those where we want fewer people on the ground: offshore sites, railway corridors, live carriageways and energy infrastructure.
3. Human observation is difficult to assure consistently
The UK's SORA framework is built around graduated levels of assurance: different levels of robustness require different levels of evidence.
Visual observation is different.
Ground observation as a tactical mitigation (M1C) exists only at low robustness in UK SORA. There is no medium- or high-robustness pathway (CAA Regulatory Library, AMC1 Article 11, M1C).
In practical terms, the regulator does not treat a person watching an aircraft from the ground as a highly reliable, measurable safety control.
Where visual mitigation is credited for air risk, the mechanism is procedural rather than performance-based. The applicant must "develop and document a VLOS deconfliction scheme, in which it is explained which methods will be used for detection", and describe phraseology "in case the remote pilot relies on detection by observers" (CAP3017, Ch. 2).
The deliverable is a document, not a demonstrated probability of detection.
And the performance of a human observer varies.
The CAA's own guidance lists the factors affecting VLOS distance: the eyesight of the remote pilot, the size and conspicuity of the aircraft, lighting, weather including haze and sun-glare, terrain and obstacles, and whether the operation is conducted during the day or at night. It concludes that "this line-of-sight distance will vary on each flight" (Sparrow Hawk VLOS Distance).
Human vision is also affected by conditions. At the limit of visibility, what is seen can become "a grayish blob that has dark and light features that are elusive and changing" (Howett, Visibility of Letters). Empty-field myopia can also affect the eye when looking into a featureless sky, with the eye resting at a distance of around 1–2 metres and the observer becoming "functionally short-sighted" (SKYbrary).
None of this means human observers are useless. It means their performance is difficult to quantify consistently.
4. It creates the wrong operating model
Visual mitigation does not scale well because its cost and complexity increase with the operation itself.
For linear assets, the arithmetic is straightforward. Under a VLOS protocol, a pilot covers roughly 3 km of a corridor at a time, requiring about five relocations to inspect 15 km (Scopito).
That is manageable for a short survey. It becomes a very different proposition when the asset is a national network stretching hundreds or thousands of kilometres.
Observer chains extend the operating range, but only by adding people at intervals along the route. As Iris Automation's Alexander Harmsen observed of waiver conditions requiring observers to "daisy chain" along a route, they "don't actually help to evaluate potential operations like long line linear infrastructure inspections; or scalable package delivery" (DroneLife, 26 Mar 2020).
More importantly, an observer-dependent model rules out the operating models that offer the greatest potential for scale.
One-to-many supervision from a remote operations centre is where the economics can improve. Zipline runs a 1:100 aircraft-to-pilot ratio with 2.3 million commercial deliveries and no injuries or fatalities (ICAO/Zipline, 2026); Wing holds 1:100 under Part 107 and 1:50 in Australia, in both cases with "no visual observers for primary operations" (Wing/ICAO, May 2026).
None of that is possible if the operating model requires one observer for every part of the route.
The UK is already demonstrating what the alternative can look like. heliguy secured UK SORA SAIL II approval in May 2026 for automated drone-in-a-box operations at two Network Rail sites, flown daily from a Remote Operations Command Centre in Newcastle, "hundreds of miles away from the operational sites" (heliguy).
And the evidence from the US suggests that observer-based approvals can become difficult to move beyond.
The US DOT Inspector General audited the FAA's BEYOND programme, which was explicitly designed to mature BVLOS. Of more than 44,000 BVLOS flights, fewer than 763, around 2%, were flown without a visual observer. Across the PSP/IPA programmes' 80,490 BVLOS flights, no operator conducted any flight without a visual observer. Only one of eight lead participants achieved scalable observer-free BVLOS (DOT OIG, 30 Jun 2025).
After years of dedicated public investment, the observer-free share remained at around 2%.
That suggests an important lesson: once visual mitigation becomes the easiest route to approval, there is a risk that it becomes the operating model rather than a temporary step towards something more scalable.
Where visual mitigation is genuinely right
I want to be precise, because overclaiming here would be unhelpful.
Ground and visual observation still has a legitimate role in several situations.
It can be a low-robustness ground-risk mitigation for occasional flights over sparsely populated terrain, exactly as M1C intends.
It can provide additional situational awareness in cluttered, obstacle-rich or unusual environments where the pilot's own view is genuinely compromised.
And it can be a mission sensor in search and rescue and emergency response, where the human is interpreting what is happening on the ground rather than being relied upon primarily to deconflict aircraft.
That last distinction matters.
An observer looking for a casualty, assessing a flood or interpreting a scene is doing a different job from an observer whose primary purpose is to detect and avoid other aircraft.
The first can add real value. The second is being asked to provide a safety function that is difficult to quantify consistently.
What is not defensible is making visual mitigation the architecture for routine BVLOS at commercial and public-service scale.
Sources
Project documents: VLOS (max) calculation JARUS ANNEX A; 20250404 – Sparrow Hawk VLOS Distance; CAP3017 UK SORA Methodology (Ch. 2, Appendices B, C, D, E); CAP3015 Detect and Avoid Policy (Ch. 1, 3, 4, 5); CAP3040 Atypical Air Environments; CAP3239B UK SORA Best Practice Guide – Atypical Air Environments; CAP3182 Future of Flight: BVLOS Roadmap; BVLOS Considerations; Drone Regulations and Net Risk; Applying Net Risk: The Safety Case for UK Drone Regulations; Howett, Visibility of Letters.
External: ATSB/BASI, Limitations of the See-and-Avoid Principle (1991/2004) · US DOT, Electronic Conspicuity White Paper (Nov 2024) · ASSURE/UND, sUAS Detect and Avoid Requirements (2017) · FAA/Volpe, Tower Controllers' Visual Detection of a sUAS (2019) · FAA/Volpe, Research Related to UAS Visual Observers (2014) · DOT OIG, FAA BVLOS Drone Operations (30 Jun 2025) · AOPA on Iris Automation Casia (Apr 2019) · DroneLife on daisy-chained waivers (Mar 2020) · McKinsey delivery-economics analysis (Jan 2023) · Kansas State University pilot-ratio cost modelling (Jan 2026) · University of Southampton, NHS drone logistics costing · Zipline, ICAO regional workshop (2026) · Wing, ICAO regional workshop (May 2026) · heliguy / Network Rail SORA approval (May 2026) · Scopito, BVLOS long-range inspection · CAA Regulatory Library, M1C ground observation · SKYbrary, empty-field myopia · National Highways GG 954, drone operations · UC Drones policy guidance, ground crew near roads · EASA Research Agenda 2026 · CAP3182 (CAA).
