A drone in a dock on a large site changes two things. It patrols the fence line without anyone walking it, and it reaches a triggered sensor in the time it takes someone to decide which camera to open.

Both of those are straightforward to set up. The part that decides whether a security programme works is the part underneath: a commercial security operator carries regulatory obligations that a police force may not, and every flight has to satisfy them before it leaves the dock.

Security is not policing, and the rules know the difference

This gets missed constantly, and it is the single most important thing for a security team to understand before buying anything.

Under Article 2(3)(a) of Regulation (EU) 2018/1139, operations carried out by military, customs, police, search and rescue, firefighting, border control or coastal surveillance services, under the control and responsibility of a member state, sit outside the EU aviation framework. Member states can choose to bring them back in, and several do, but the starting position is that a police drone programme and a commercial one are governed differently.

A security company protecting a private site is a civil operator like any other. Flights beyond the pilot's visual line of sight put you in the specific category, which means a risk assessment under SORA 2.5 and an operational authorisation from your national aviation authority. The conditions in that authorisation then apply to every single flight, including the automated ones.

So a security buyer is really asking whether the system makes it practical to fly the same way every time, in a form you can show an auditor.

Patrolling the fence

A perimeter patrol is the same route, flown repeatedly, which is exactly what a saved mission is for. Build the route once in the mission editor, set the height, the camera angles and the points where the aircraft should hold, and the patrol runs identically whoever is on shift.

That repeatability matters for two reasons beyond convenience. Comparing today's pass against last week's only works if the aircraft flew the same line at the same height. And an authorisation granted for a defined operation stays valid only while the operation stays the same, which a saved mission enforces and a manually flown patrol does not.

Responding to a sensor that has no API

Most perimeter sites already have sensors: fence vibration, beam breaks, cameras with analytics. The ones installed years ago often have no interface for anything to connect to.

There is a practical way around that. Build a template per sensor, named for the sensor it covers, with a route from the dock to that sensor's position. When Sensor 3 trips, the operator runs the Sensor 3 template and the aircraft flies there.

The routing is handled for you. Zones and no-fly zones defined on the map are respected on the way, so an aircraft heading to a sensor near a boundary routes around the area it may not enter instead of taking the direct line. Map markers let you label what matters on site: gates, hazards, the places an operator needs to find quickly in the dark.

Where a sensor system does have an interface, the same response can be triggered automatically. The template approach is what makes the other ninety per cent of installed estate usable today.

Satisfying the authorisation before every flight

This is where an operations platform earns its place for a security operator.

The checklist. Your SORA produces conditions, and those conditions have to be met before each flight. A checklist built into the launch sequence turns that from a document in a drawer into something the operator works through on screen, with a record afterwards of what was checked and by whom.

Weather at the dock and in the forecast. Conditions at the dock itself and the forecast for the window ahead both feed the go or no-go decision, and for most authorisations that decision has to be documented.

Traffic in the airspace around you. The SafeSky integration shows crewed traffic at low altitude, aggregated from more than thirty sources including ADS-B, Mode-S and FLARM, alongside a community of general aviation pilots. For a security operator this is more than awareness. Real-time traffic data is recognised as an air risk mitigation in a SORA, and operators have used it to argue a lower air risk class in their assessment. It is worth raising with whoever writes yours.

What the operator actually works with

Payload choice follows the task: a wide camera for the patrol, thermal for anything at night or anything involving a person who does not want to be seen. Both feed the same live view.

When the automated route stops being the right answer, the operator takes manual control with keyboard or joystick and flies it themselves. 

The feed can also go where the team already looks. AirHub brings drone video into a video management system through streaming protocols including RTSP and RTMP, so aerial footage appears next to the fixed cameras in the control room instead of on a separate screen. The Drone Operations Centre works the other way too, pulling feeds, maps and data into one shared picture for whoever is coordinating.

Where the data lives

Perimeter footage of a client's site is sensitive, and for many security contracts the question of where it is stored arrives before the question of what the system does.

AirHub is a Dutch company. The platform is ISO 27001 and ISO 9001 certified, operates in line with GDPR and European data protection requirements, and can be deployed on-premise or in a private cloud where an organisation needs full control over its operational data. For a security provider answering a client's security questionnaire, that is usually the section that takes the longest.

Our security page sets out how the pieces fit together, and our regulatory consultancy team works with operators on the authorisation route itself.

DJI launched Dock 4 today, with a new aircraft, the Matrice 5TD. The specification sheet is long, and one number on it matters more than the rest: the Dock swaps a depleted battery for a charged one in 25 seconds, where every Dock on the market until now recharged the pack inside the landed aircraft and took around half an hour to do it.

That single change rewrites the arithmetic behind a dock network. Here is what it means for the three kinds of operation we work with most.

What 25 seconds changes about coverage

A dock that recharges in place covers its radius for part of the day. Respond to one emergency, return to the Dock and the drone is not available for the next thirty minutes, which is the window in which the second call arrives. Programmes have worked around this by buying a second dock, which means a second aircraft, a second roof, a second power supply and a second set of permissions, to solve a problem that was fundamentally about one flat battery.

With a 25-second turnaround, availability stops being the limiting factor and distance takes over. The useful calculation becomes simple. The Matrice 5TD cruises to a scene at up to 21 m/s, so a minute of transit covers around 1.25 km in a straight line. Subtract your launch sequence, subtract climb time, and add nothing for the route being direct, because no-fly zones and obstacles make every real route longer than the line on the map. What you are left with is your true coverage radius, and with a swap-equipped dock you hold it continuously instead of for half of each hour.

Flight endurance moves too. A single Matrice 5TD flight runs up to 59 minutes, so a longer sortie is followed by a short pause rather than a long one.

Sergey Georgiev, AirHub´s Senior BD Manager and ex-DJI:

“Every public safety drone team I speak to asks the same question about the gap between flights. Until today the only answer was a second dock. Additionally, one little change will improve operations even further: the increased horizontal speed of 21m/s will allow DFR to be even quicker on site.”

Public safety: the gap between the call and the picture

For a Drone as First Responder programme this is the constraint that mattered most. Dubai Police already reaches scenes in around 90 seconds with AirHub, and the published study of their 2023 operations records mean arrival times of 1 minute 25 to 1 minute 28 across 1,502 incidents.

Two other details help here. Drone-to-cloud latency on the new setup is up to 160 ms, roughly a third lower than the previous generation, which shows up as a smoother live view for whoever is making decisions. And full-colour 4K night vision without a spotlight changes what a night-time scene assessment looks like, with an infrared camera alongside it for anything thermal.

The aircraft also carries payloads up to 450 g, including speakers and spotlights. For a search at night, a speaker that can address someone on the ground is a different capability from a camera that can see them.

Critical infrastructure: inspection without swapping airframes

The Matrice 5TD combines telephoto, infrared, laser ranging and night vision on one aircraft, which removes the usual trade-off of choosing a payload before you know what you will find.

The 11× telephoto resolves detail at 400 m and picks up pin-level defects on a power line from 100 m. The infrared camera reveals overheating and damaged insulators. Laser ranging works to 3 km with coordinates overlaid on the feed. Survey accuracy runs to 10 cm.

The obstacle sensing is what makes this usable around live assets. Three-directional mmWave radar, a rotating LiDAR and six-directional vision sensing work together, and the wire-level detection picks up thin obstacles including power lines at night.

Security: overnight patrols that actually run overnight

For perimeter work the charge cycle set the patrol interval, which is a poor way to decide how often you check a fence. Remove it and the interval follows the risk assessment instead.

A few things in this release are aimed squarely at security operations. The built-in customisable warning light works during patrols with nothing to install. Speakers and spotlights are available as payloads. The Dock operates from -40°C to 50°C, preheating batteries in freezing conditions and running air conditioning in heat, and the aircraft takes off, lands and operates in winds up to 12 m/s, with an automatic descend-and-return behaviour if wind spikes while it is out.

Weather is the single largest cause of missed patrols and missed response targets, so a wider operating envelope is worth more than it sounds.

The regulatory detail worth more than it looks

One line in DJI's announcement deserves more attention than it will get. The Matrice 5TD carries a Flight Termination System, which DJI describes as a key technical requirement for C6 compliance, triggered manually or automatically when the aircraft breaches a geocaging boundary.

C6 is the class marking required for STS-02, the European standard scenario for flights beyond visual line of sight. An operation that fits STS-02 runs on a declaration to your national aviation authority rather than a full operational authorisation, and that declaration is recognised across every EASA member state. For a European operator, the difference between the declaration route and a bespoke SORA can be months.

Whether a specific aircraft and configuration actually carries C6 class marking is a question for the manufacturer and your authority, and it is worth asking early, because it shapes your whole timeline. Our regulatory consultancy team works through exactly this question with operators.

Where AirHub fits

A dock that is ready all shift flies far more often, and every one of those flights still needs a route planned, someone watching the feed and a record afterwards. That workload grows in step with the flight count, and it is the part AirHub takes on.

Sergey:

“The increased horizontal speed of the new Matrice 5TD combined with AirHub's SmartDispatch and intelligent route planing allow Public Safety entities to be on site even quicker than before.”

Mission planning and saved routes in the mission editor, so a response or a patrol flies the same way every time and one round compares against the last. Live operations and a shared picture in the Drone Operations Centre, so the people deciding see what the aircraft sees. Fleet management across your aircraft, docks and pilots. And a flight record that holds up when an authority asks for it.

AirHub is hardware agnostic by design, which matters more as hardware cycles get faster. A platform tied to one manufacturer ties your fleet decisions to that manufacturer for the life of the programme.

AirHub is also a Dutch company. The platform is ISO 27001 and ISO 9001 certified, operates in line with GDPR and European data protection requirements, and can be deployed on-premise or in a private cloud where an organisation needs full control over its operational data.

Our public safety and security pages set out how the pieces fit, and we will be showing the new Dock in action with AirHub shortly.

Claims about Drone as First Responder are easy to find. Published evidence is harder. Most of the numbers circulating in the market come from vendor pages and conference slides, and a procurement committee will treat them accordingly.

That picture changed a little in September, when a study of one of the largest programmes in service appeared in a peer-reviewed journal. This post sets out what has actually been measured, where the figures come from, and which claims still have nothing behind them.

The strongest dataset so far

In September 2026, Policing: A Journal of Policy and Practice, published by Oxford University Press, carried a study of the Dubai Police Drone Box system, written by researchers at the Dubai Police Academy together with the force's General Department of Operations.

The study covers the whole of 2023 and every drone-assisted response in that period: 1,502 incidents, of which 1,496 were classified by severity. That matters methodologically, because it is the complete population and not a sample, which removes the question of how cases were selected.

The response times are consistent across incident types. Mean arrival was 1 minute 25 seconds for incidents triaged as very critical, 1 minute 28 for critical, and 1 minute 26 for normal. Medians sit within a second or two of the means, standard deviations stay under 11 seconds, and the slowest single response recorded was 2 minutes 6 seconds.

The measurement starts at the trigger, which is the emergency call, a CCTV detection or a system alert, and ends at confirmed arrival on scene. Timestamps are logged automatically by the platform, so the figures do not depend on anyone remembering to write a time down.

The consistency is worth more than the average. A system landing between 1:00 and 2:06 across fifteen hundred incidents is behaving predictably, and predictability is what an operations manager plans around.

The operation runs beyond visual line of sight from a central control room, around the clock, under a specific operations risk assessment submitted for each location and licensed by the Dubai Civil Aviation Authority.

The scale is worth knowing, because it puts the incident count in proportion. In the period studied the programme ran from eight Drone Boxes plus one mobile unit, with planned expansion towards 800 square kilometres of coverage. So 1,502 responses in a year came from a network of that size, which is a more useful benchmark for anyone sizing their own programme than the response time on its own.

What American programmes publish about themselves

Two US forces publish running figures, and the Dubai study cites both.

Brookhaven Police in Georgia reported responding to 1,700 calls for service in 2024, with the drone first on scene for 72 per cent of them and an average response time under two minutes. Chula Vista Police in California, the longest-running programme of its kind, publishes a live dashboard: between 1 January and 28 October 2025 the force responded to 2,317 calls, the drone arrived first at 943 of them, average response time was 2.4 minutes, and the programme avoided dispatching 109 patrol units.

Both forces publish their own operational data and keep it updated, so anyone can check the current figures directly.

The medical parallel

The closest thing to a European controlled comparison comes from emergency medicine. A Swedish study published in Resuscitation in 2024 compared drones dispatched to time-critical incidents against ground EMS. Across 52 cases, and despite unfavourable weather and closed airspace in some of them, median drone response was 3 minutes 49 seconds against 5 minutes 51 for the ambulance, and the drone arrived first in 47 of the 52.

Slower than the policing figures, in a different operating model, and peer-reviewed. For anyone arguing a case to a sceptical board, a published European study showing the same directional result is worth having alongside the policing numbers.

What has not been established

Four things are missing from the literature.

Cost savings are not quantified. The Dubai authors name this explicitly as missing: nobody has yet calculated the financial comparison between sending a drone and sending a patrol officer, including the time lost when an officer attends an incident that did not need them. Every euro figure you will see on this subject is currently a vendor's model.

The effect on crime is unmeasured. Changes in crime types and rates before and after a drone programme would be valuable, and the study lists it as future work.

Nothing is longitudinal. All the strong data is a single year or less, at a single site. Trends over time, and whether the early performance holds as a programme scales, remain open questions.

Call-clearance rates vary wildly. The share of calls resolved without a ground unit is reported anywhere from about a fifth to more than half, depending on who is publishing and which calls are counted. Chula Vista reports 109 patrol units not dispatched over its ten-month period, and that figure counts units and not calls, so it does not map onto those percentages at all.

What to do with this

Response time is the claim that holds up. Sub-two-minute arrival is documented across three programmes on two continents, one of them in a peer-reviewed journal with a full year of data. That is the number to quote, with the source attached.

Everything financial is still unmeasured. No cost comparison between a drone response and a patrol response has been published anywhere, so any figure on this subject is a model and not a finding.

The gap is European data. Every policing dataset here comes from Dubai or the United States. No European force has published comparable response-time figures, which means operators here are currently arguing from other people's jurisdictions.

A last point on Dubai. The published study documents the performance of the Drone Box system as operated by Dubai Police, and it does not assess the software behind any particular deployment. Dubai Police is an AirHub customer and we are proud of the work, and the study stands on its own as evidence about what a mature dock-based programme can achieve, which is how it should be read.

If you want to talk through what this means for a programme in your own jurisdiction, our public safety page sets out how the platform supports each stage, and our regulatory consultancy team works with operators on the authorisation route.

*Picture: Dubai Police Drone Box (Source: Dubai Police UASC 2025).

The control room decides whether a Drone as First Responder programme becomes routine. The aircraft and the dock are the parts people photograph, and the part that determines daily use is whether the drone fits the systems a dispatcher already has open.

This blog covers how that connection works: what computer-aided dispatch (CAD) does, where a drone connects to it, how aerial video reaches the screens your operators already watch, and what to ask your existing providers before you commit to anything.

What CAD is doing, and where the drone fits

Computer-aided dispatch is the system that takes an incoming emergency call, turns it into a structured incident record with a type, a location and a priority, and assigns resources to it. Everything a control room does flows through it.

A DFR programme attaches to that flow at one point: the moment an incident is created and classified. That record already contains what a launch decision needs, which is a location and an incident type. So the integration question is narrow. Can the incident record reach the drone system quickly and reliably enough to launch on it?

There are three ways that connection is usually built, and they differ in how much authority the CAD system hands over.

Triggered from dispatch. The incident record reaches AirHub directly, and an incident of a qualifying type starts the DFR workflow: the best-positioned dock is selected, a route is calculated around weather, airspace and no-fly zones, and the aircraft launches under operator supervision. This is the version that delivers the response time, because nobody retypes anything. 

Dispatcher-initiated, CAD-informed. The incident appears in the drone system alongside the dispatch record, and an operator confirms the launch with one action. Slower by seconds, and often the right starting point, because it lets a control room build confidence before handing over the decision.

One-way, after the fact. Flight data flows back into the incident record for reporting and review, without CAD triggering anything. Worth building early regardless of which of the other two you choose, because it is what makes the record complete. 

Getting the picture onto the screens people already watch

The second half of integration is video, and this one is better standardised.

AirHub brings drone footage into a video management system (VMS) through streaming protocols including RTSP and RTMP, which most control room video environments already speak. That means aerial footage appears alongside the fixed camera feeds, on the same screens, which is what keeps a drone programme in daily use. 

For an organisation running a real time crime centre, that is the point where drone coverage becomes part of the existing picture: fixed cameras cover what they cover, and the aircraft covers the ground between (and above) them.

Video travels the other way as well. Where the paragraph above is about sending drone footage out to your video management system, the Drone Operations Centre pulls feeds in: multiple live video sources, maps and data in one shared view, so that operators, commanders and, where appropriate, external agencies all work from the same picture. Counter-UAS sensors feed into the same view, so activity in the airspace around an incident sits next to activity on the ground.

What has to be true on your side

Integration work is usually less about the drone system than about four things in your own environment.

The incident record has to carry a usable location. A structured coordinate or address field, not free text in a notes box. This is the most common blocker and it is worth checking before anything else.

Incident types have to be classified consistently. Automatic launch depends on a rule, and a rule depends on your dispatchers coding incidents the same way. If classification varies by shift, fix that before you automate it.

Somebody has to own the interface. CAD environments in public safety are frequently customised, sometimes by a third party. Find out early who is allowed to change yours and what their timeline looks like, because it is regularly longer than the drone procurement.

Security review takes as long as the build. Anything touching dispatch goes through information security, and in a public sector organisation that is a process with its own calendar.

What to ask your CAD provider

Five questions, in the order that will save you the most time.

1. Does the system expose incident events through an interface, and is it documented? 

2. Can a third party subscribe to incident creation, or does every integration go through you? 

3. What does a change request cost and how long does it take? 

4. Are there existing drone or field-system integrations we can look at? 

5. And what are the security requirements for an external system receiving incident data?

Ask these before selecting software. The answers routinely change the shape of the programme, and occasionally the order in which it gets built.

Where the data goes

Anything touching dispatch data raises the question of where it is processed and stored, and in public safety that question arrives early.

AirHub is a European company. The platform is ISO 27001 and ISO 9001 certified, operates in line with GDPR and European data protection requirements, and can be deployed on-premise or in a private cloud where an organisation needs full control over sensitive operational data. For an integration touching live incident records, that last option is often what makes the security review straightforward rather than lengthy.

Where to start

The sequence that works is a slow one. Confirm what your CAD environment can expose and who can change it. Decide which incident types would justify an automatic launch. Start with dispatcher-initiated launch, so operators build confidence before the system takes the decision. Get video into the VMS early, because it is the part that makes the programme visible to everyone else. Then automate the trigger once the rest is routine.

Our public safety page sets out how the pieces fit, and we are happy to go through what integration would involve in your environment.