Airffic InstituteAERIAL PHYSICAL AI

Aerial Physical AI Institute

Autonomy: how a drone decides and acts

After seeing and understanding comes deciding. Decision-making sets how wide a drone's job can be.

Short answer: The value of autonomy is how much human supervision it removes. Moving from obstacle avoidance to detect-and-avoid (DAA) to many aircraft at once lets one pilot oversee more drones — and that changes the business case.

Updated: · 3 min readHow we researched →

Sir Airffic GoIt looks like a technology story, but it is really about money: how many drones can one person run?

Levels of autonomy: how much to hand to the machine

LevelMachine doesHuman doesAircraft per person (indicative)
1 ManualStabilises attitude onlyAlmost all piloting1
2 AutomatedFollows a set routeWatches and decides1
3 Obstacle avoidanceAvoids nearby objectsMission decisions, monitoring1 to a few
4 AutonomousReplans and completes the missionHandles exceptions onlyA few to many
5 Swarm & airspaceDrones and airspace OS coordinateSets overall policyMany

The per-person figures explain the mechanism; they are not regulatory limits. Real limits come from national rules and operational approvals.

Obstacle avoidance: the first step to not crashing

Around 2016, consumer drones began to detect obstacles ahead and stop.

Then drones with all-around cameras appeared that avoid obstacles on their own while following a person (Skydio and others). Even a beginner could no longer easily crash them.

Obstacle avoidance is now expected in commercial drones. The difference lies in how well a drone copes with hard cases such as thin wires and night flight.

Detect-and-avoid: not hitting planes and helicopters

Detect and Avoid (DAA) means spotting crewed aircraft and getting out of their way. It is essential for flying beyond visual line of sight (BVLOS).

Compact radars and camera-based AI that spots distant aircraft have been developed to watch the sky instead of a person.

Regulators treat DAA performance as one condition for allowing BVLOS. DAA is both a technology and a regulatory key.

Flying in swarms: swarms and defense

A swarm is many small drones that communicate and move as one group.

In 2022 a Zhejiang University team published research on swarms of small drones flying autonomously through a forest.

Around the same time, the war in Ukraine sharply increased investment in drones that fly the final stretch on their own despite jamming, and in letting a few people run many aircraft. Defense demand is also accelerating civilian autonomy.

Sir Airffic GoDefense cannot be ignored. But this site explains mechanisms and markets; it does not teach how to build weapons.

Autonomy and regulation move together

However smart a drone is, it cannot fly commercially unless the state allows it. Japan created Level 4 (BVLOS over people) in 2022.

In the US, the Part 108 proposal to allow BVLOS under standard rules was published in 2025. Europe is building U-space for airspace management.

Each regulatory opening raises the value of autonomy. Strong companies invest in dialogue with regulators as much as in technology.

Frequently asked questions

What is BVLOS flight?

Beyond visual line of sight: flying a drone farther than the operator can see it. It is needed for delivery and wide-area inspection and requires special approvals or rules in most countries.

What is detect-and-avoid (DAA)?

Technology that detects other aircraft and avoids collisions. It uses compact radar or camera-based AI and is a key condition for BVLOS approval.