Airffic InstituteAERIAL PHYSICAL AI

Aerial Physical AI Institute

Twenty years in four eras, and the turning points

Today's market has reasons for its shape. Split 20 years into four eras and the reasons connect.

Short answer: Aerial Physical AI moved from the lab (2006–) → the camera boom (2013–) → enterprise and regulation (2017–) → defense and foundation models (2022–). With each era, what counted as valuable moved from the aircraft to scale, to trust and compliance, to autonomy and origin.

Updated: · 2 min readHow we researched →

Sir Airffic GoHistory is not for memorising. Follow why each era changed, and you can read the next ten years.

Four eras

2006–2012Lab eraUniversities estimateposition from cameras andIMUs2013–2016Consumer camera eraDJI wins at scale;obstacle sensors arrive2017–2021Enterprise & regulation eraInspection and deliverystart; the US moves awayfrom DJI2022–2026Defense & foundation-model eraWar accelerates autonomy;AI that sees, reasons andflies
  1. 2006–2012 Lab era
    Universities estimate position from cameras and IMUs
  2. 2013–2016 Consumer camera era
    DJI wins at scale; obstacle sensors arrive
  3. 2017–2021 Enterprise & regulation era
    Inspection and delivery start; the US moves away from DJI
  4. 2022–2026 Defense & foundation-model era
    War accelerates autonomy; AI that sees, reasons and flies

Era 1 (2006–2012): the lab

The drone's brain lived in university labs. Research on estimating a small aircraft's position with a camera and an IMU appeared one after another.

Videos of fast, precise flights spread worldwide and created the idea that drones can fly themselves.

Algorithms published in this era became parts of later products. Open research laid the market's foundations.

Era 2 (2013–2016): the camera boom

China's DJI mass-produced ready-to-fly camera drones and quickly took the world market.

Amazon announced its delivery drone plan and Google began delivery tests. Expectations grew that the sky would become a workplace.

Around 2016 obstacle sensors arrived on consumer aircraft. The brain moved from the lab into mass products.

Era 3 (2017–2021): enterprise and regulation

Inspection, survey and delivery use took off. In the US, Part 107 (2016) set rules for commercial use, and delivery firms began to obtain air carrier certification.

Meanwhile, security concerns led the US to tighten restrictions on DJI, and money began to flow to non-Chinese aircraft and autonomy software.

The value of this era was trust: safe flight, protected data and compliance became reasons to buy.

Era 4 (2022–2026): defense and foundation models

The war in Ukraine made demand for cheap autonomous drones and counter-drone technology explode. Huge sums flowed to defense start-ups.

In 2023 an AI drone beat human racing champions, and foundation models began serving as robot brains.

The US published new BVLOS rules and executive orders backing the drone industry. Autonomy, regulation and security now move at the same time.

Sir Airffic GoThe fourth era is not over. We expect the next protagonist to be systems that run many smart aircraft safely across the whole sky. That is our interpretation; the evidence is in the news timeline.

Frequently asked questions

When did research on autonomous drone flight begin?

From the late 2000s university labs intensified research on camera-and-IMU self-localisation and autonomous flight of small drones; from the mid-2010s obstacle avoidance began to appear on consumer aircraft.