Safety of outdoor drone shows in public spaces
Watch an outdoor drone show and the air seems to have turned soft and obedient. Points of light appear, gather into shapes and vanish again, as calmly as if this were simply an animation running on an enormous screen.
But air is not a screen. And a city is not a studio.
In public spaces, safety is never a separate subject. It does not begin at the moment the drones are already in the sky. It begins long before that, with how we mark out the boundaries, how we build the perimeters, how we survey the site, how we run the tests and how we make decisions when conditions change.
We never fly over people
The first rule is simple and it is hard. We never work above people.
Not because drones are bound to fall. Because in public projects we leave no room even for the unlikely scenario. Where the drones take off, where they transit to the display point, where the show itself takes place and where they return, there must be no people. That logic sets the whole geometry of the project. Until the perimeter is closed and the zone is cleared, the drones do not take off.
So we always fence off the launch area and the technical area physically. This is not improvisation; it is a standard in our technical rider. Inside the perimeter there is only the working crew. Outside the perimeter there is security. If it is a city event, the municipal services and the police join the operation. And it matters to us that a security representative is always beside the team, with a direct radio link. In a public space, safety is always teamwork and the equipment is only one part of it.
Around the launch area we usually keep a buffer of some 50 metres from the drones laid out on the ground. The exact figure depends on the site, but the principle does not change. There has to be empty space around the drones, to rule out anything straying in, whether people, vehicles or crew and to keep the return zone controlled.
Distance to the audience is set by the picture, not by safety
We are often asked about the distance to the audience. And here it is important to separate two different things: safety and perception.
Safety, for us, is settled by the perimeter and by the fact that there are no people under the drones. The distance to the viewer is more often decided by the geometry of the picture. In a show of 300 to 400 drones the aerial stage is usually somewhere between 60 and 120 metres across and the same in height. A picture like that is pleasant to watch from closer in, sometimes from 50 to 150 metres. In a show of 1,000 drones or more the stage can reach 200 metres across and 200 metres high, plus depth if it is in 3D. That is a flying volume, almost a stadium in the air and it is more comfortable to watch from further away, 200 to 250 metres. We work with the client to set a distance that keeps the scale of the show readable, rather than cutting the distance down for the sense of being close.
Altitude follows the space, not the specification
Flight altitude does not lead a life of its own either. Technically the drones can fly from 5 metres up, but in real public projects we always start from the task and the space. More often that means altitudes from 20 metres. At the opening of the light festival in Sharjah, for instance, we began the main animation at 27 metres, because in front of the audience stood a 25-metre building carrying a mapping projection. We raised the lower boundary so that where the building ended, the sky began for the viewer and with it the drone show. It is a simple thing, but predictability is built out of details like that.
Fault tolerance is architecture, not reaction
After that comes the part that is almost invisible from the outside. Fault tolerance.
In the air you cannot count on everything being perfect. So we structure the work so that the system does not lose control, even when something changes.
It is important to understand this: we do not build the system to react heroically to a failure. We build it so that most critical situations never reach the stage of an incident. Fault tolerance is not a reaction; it is an architecture.
If a drone loses GPS, it does not do anything abrupt. For about two minutes the on-board software tries to recover the signal. If it cannot, the drone automatically returns to the launch zone, to the point where landing begins and holds there at a working altitude, 50 metres for example, waiting for the others to finish the sequence and come back. It does not land on its own somewhere off to one side. The logic is precisely this: not to create unexpected landings outside the controlled zone.
If the control link is lost, the logic is the same. What matters to us is that the behaviour of the system stays predictable: it has lost its references and it returns to a zone that was cleared and fenced in advance.
The chief pilot, and control drone by drone
We have automatic modes and we have human control. Return to Home exists. But the key point is that the chief pilot can stop the show at any moment, in one of two ways: land the drones immediately, or stop the sequence and bring them back to the launch zone. The chief pilot does not simply start the animation. His job is to keep safety, the air and the real environment in mind, rather than the picture on the screen.
And there is one more important tool that is rarely talked about publicly. We can remove any single drone from the sequence manually. Even with a thousand aircraft in the air, we can see one particular drone, select it and land it separately. Which means the system is not monolithic. It is controlled point by point. And if one element is behaving unreliably, we do not risk the whole show to get as far as the final scene.
The checks start long before the show
For all of this to work, we begin checking long before the start.
On the day of the show we never rely on a feeling that everything is fine. We run tests. If it is possible to get clearance to fly the day before the event, we run several rehearsals in advance. The first compulsory step is the hover test. The drones take off in layers and hold position while we check the link, the stability, the wind and the general state of the system. The hover test is compulsory before any show. It is a basic procedure and we do not begin a display without it.
If instability shows up at this stage, whether in the link, in position holding or in how the system responds, the show is postponed or adjusted. We do not hope that it will get better up there. We register the problem before the start.
After that, if the schedule and the permits allow, we run a rehearsal show. Sometimes complete, sometimes partial. We do not have to put all the thousands of drones up to check the geometry. We can take particular sequences out of the animation, the extreme points for example and put up a notional 50 drones to see how the boundaries of the stage behave, how the link holds and how precisely the shape assembles. That saves wear on the equipment and gives a real check rather than a theoretical one.
The radio environment and the ground it starts from
A separate layer is the radio environment and the launch site.
We always check the radio environment in advance. But not only the radio environment. We check the site for everything that might interfere with the link and the stability of the system. We walk the launch zone with a dedicated instrument and look at the geomagnetic readings. It is important to us that there are no sources of interference under the drones, no reinforced concrete ducts, no cabling, no metal elements, nothing that could degrade signal transmission or create instability at the start. This sounds like a small thing until you come up against a real city site, where a beautiful square may have dense infrastructure underneath it.
The main control channel by default is Wi-Fi. But the system is backed up by a radio channel. If Wi-Fi is lost or degrades for any reason, the switch to the radio frequency happens automatically, in a fraction of a second. A switch back is possible afterwards. This is not manual magic and it is not luck. It is built-in redundancy logic.
Dense urban development brings additional difficulties. It is not that we would be unable to fly. Trajectories and zones can be calculated. The hardest things in a city are link noise, signals reflecting off buildings and the quality of GPS. Which is why city projects call for even more careful radio preparation and for control throughout.
Antennas matter here. Ours are directional, and we turn them towards the drones. We follow the swarm from take-off to landing, aiming the antennas at wherever the drones are and keeping signal transmission reliable and precise. Communication in a show is not a setting made at the beginning. It is a process that runs the whole time the drones are in the air.
We never judge wind by how it feels
Now the weather. The very factor that many people would like to treat as background.
We never judge wind by how it feels. Wind can differ at different altitudes, and that is critical. Down low it may be 6 to 7 m/s, while at 200 to 250 metres it is already 13 to 15 m/s. We measure the wind in the zone where the flight is actually planned, because wind has one important property: it changes. And that variability is often zonal, so conditions can differ by altitude and between points on the same site. Which is why, before a launch, we always know the size of the animation and the minimum working altitude of the drones. If gusts at the animation altitude exceed roughly 12 m/s, flying is not safe. Controlled flight in general is possible up to 10.8 to 12 m/s, about 39 to 43 km/h, but what matters most is the gusts and the wind profile with altitude.
And here the role of the chief pilot comes in again. He can cancel the show, postpone it, move it earlier or later, working from the forecasts of dedicated systems and from the actual conditions. With the organisers’ agreement we have moved a show several hours earlier, because the forecast showed the weather deteriorating substantially by the planned time. There have been cases when we moved a show to the next day because of storm-force wind. And there was a situation when the wind began to build during the show and the chief pilot decided to bring the drones home with only one picture left before the finale. In our work the final scene is never more important than control.
If even one key parameter goes outside its limits, whether in the wind, in the link or in the regulator’s clearance, the show does not start. Even if the site is ready, the audience is in place and the timing has been announced. Safety is not agreed after the fact. And the size of the audience changes nothing here. On projects with tens of thousands of spectators the standard stays the same: a cleared perimeter, a controlled flight zone, an isolated return zone. The more people are on site, the stricter the geometry of safety.
Rain has an effect too, but a different one. First of all it reduces the clarity of the air; light scatters, contrast drops and the show reads less well. That is a question of the result, not only of technical capability. Snow and wet snow are more dangerous, because they can stick and change the aerodynamics. Any difficult weather increases the load on the system and shortens flight time, because the drones have to spend more energy holding position and working against the wind.
The operating temperature range of the drones is –10 °C to +50 °C, but it is best to stay below +45 °C. Batteries do not like overheating. Above roughly +45 °C the risk of cells overheating and of the battery failing grows. We watch the state of the batteries constantly, because this is not simply a question of whether there is enough energy. It is a question of safety.
On charge level, lithium-ion batteries give a simple guide. The safe minimum is 20 per cent. Going below 10 per cent can destroy a cell and carry a risk of ignition. So we do not fly in a way that squeezes a battery to its limit. We plan the flight window with a margin, particularly when conditions are difficult and energy consumption rises.
And there is one more factor that is often overlooked: the airspace itself.
Even if the equipment is ready, take-off may be forbidden. The civil aviation services, the Ministry of Defence or other authorised bodies can restrict flights at any moment. Then the drones do not fly. Full stop. And the decision about what to do next lies with the chief pilot again: wait, postpone, shift, cancel. In the air you do not argue with the regulator.
From our real cases there is one that shows very well why all these layers of control are needed. On one project the state services did not switch off some of the GPS jammers. As a result, at an altitude of around 250 metres, the upper drones began to drift. The system worked as designed: the instability did not turn into a loss of control and did not require an emergency landing. It was unpleasant, but it did not become a disaster, because our animation never rose above 300 metres and we kept the situation under control. Moments like that make it clear that reliability is not a promise but an architecture: redundancy, procedures and a pilot who knows when to make the call.
The safety of an outdoor show is never built on a single decision. It is always several layers: the ground, where the perimeter is controlled; the air, where the geometry of the flight is calculated; the link, where the channels are backed up; the energy, where the margin is set; and the person who makes the final decision. If even one layer is unstable, the show does not begin.
In short, the safety of an outdoor drone show in a public space looks like this.
First we clear the air and the ground, and only then do we put the drones up.
First we check the site, the link and the stability of the system and only after that do we begin the show.
First we build in the margin on battery and weather, and only then do we settle the length and the final structure of the animation.
The audience sees lightness in the sky.
And our work is to make sure that behind that lightness there are always boundaries, control and calm.
