Post
Conversation
![David Watson ๐ฅ](data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" width="96" height="96"><rect fill-opacity="0"/></svg>)
Post your reply
The technical breakthrough wasn't just academic theory - it addressed the problem helicopters being fundamentally unstable.
Like trying to balance a pencil on your fingertip while walking, helicopters require constant corrections to stay airborne.
Only military applications
Show more
0:03 / 0:35
Traditional helicopter control required unique mathematical models custom-built for each aircraft.
Imagine needing separate software for every smartphone model instead of a universal OS.
This fundamental problem had to be solved for commercial drone adoption to become possible.
Wang's key insight? Stop trying to create perfect mathematical models.
Instead, design adaptive feedback loops using real-time sensor data to make continuous micro-adjustments.
This shift to practical adaptation formed the cornerstone of DJI's future technical advantage.
The system Wang designed checks position 50 times per second and makes immediate corrections.
No complex physics calculationsโjust continuous micro-adjustments based on sensor feedback.
This approach resembles how we intuitively balance on bicycles without conscious thought.
DJI's journey was further bolstered by miniaturized gyroscopes, accelerometers, barometers, compass, and GPS.
Strategic component positioning minimized vibration and electromagnetic interference from motors.
This integration became a key competitive advantage.
Wang's control architecture featured three interlinked systems:
Hovering control (maintaining position)
Semi-auto flight (velocity commands)
Ground station navigation (following waypoints)
See how this approach using nested PID control loops created DJI's reliability edge?
The real-world results were impressive even by today's standards:
Hovering accuracy within 0.18m
Navigation over 7.8km courses
Velocity deviations under 0.25m/s
These metrics validated that the approach could work reliably in commercial products.
This research emerged parallel to DJI's founding and growth under Frank Wang and Professor Li Zexiang.
Not coincidentalโthere is a deep technical foundation that has led to the consumer products.
Within years, DJI captured 70% global market share using these principles.
Safety features pioneered in this research also became DJI's competitive moat:
Fail-safe protocols for communication loss
Indoor stabilization without GPS
Vibration isolation for sensor protection
These now-standard features were crucial when introduced to consumer markets.
Perhaps most innovative was the auto-tuning system using frequency analysis.
Similar to how smartphones auto-adjust camera settings, the control system could calibrate itself.
This allows DJI drones to adapt to different aircraft characteristics without user intervention.
Today's applications weren't possible before these control systems made autonomous flight dependable:
Real-time mapping
Aerial photography
Inspection of infrastructure which previously required helicopters
DJI's 15-year R&D head start built directly on this research.
DJI also leveraged these principles to develop:
RTK modules for precision positioning
Advanced mapping and surveying capabilities
Efficient data processing requiring minimal computational resources
This technical foundation drives DJI's continued market leadership.
Manufacturing expertise amplified the technical advantages.
DJI leveraged China's dominance in plastics, small electric motors, and high-volume electronics production.
The combination of advanced control systems with manufacturing scale created a dominant market position.
0:27
When did drones transition from specialist military tools to everyday consumer devices?
This research marks that inflection pointโwhen specialized expertise was no longer required.
Similar to how graphical interfaces democratized computing beyond programmers to everyday users.
Urban air mobility and flying taxi projects now build directly on these control principles.
They've evolved with better hardware and additional redundancies for human transport.
Today's aviation innovations stand on similar research foundations.
The most successful technologies become invisible infrastructure we take for granted.
That's why DJI succeededโmaking complex technology fade into the background.
What began as a research project on helicopter stability created a new category of accessible flying devices for
Show more
More insights on AI, robotics and aviation at
Quote
Ashish Kapoor
@akapoor_av8r
Drone giant DJI didn't appear overnight.
Their decade-long dominance stems from CEO Frank Wang.
Wang's research and his eventual 2011 thesis defined the blueprint that revolutionized quadcopters forever.
Let's revisit this thesis here: 
Show more
one of djiโs first drones from 2012, the flamewheel F550 hobby kit. figuring out how to build this in high school and fly it around the neighborhood is a core memory. pissed off so many of neighbors with this thing hahaha
Anyone know why I can't buy a DJI Neo or a DJI Flip on the Best Buy website right now. There's been no listings or inventory for months. Only expensive $2k+ DJI drones are available.
I went from KK2 flight controller to Naze32 and then I bought a Phantom 2. Wow what a leap forward in terms of ease of use - no more tweaking gains and experimenting. Hovering my homemade quads in mild winds required constant adjustments in-flight.
Long awaited thread! Thank you for sharing this insightful thread. If possible, could you please also explain a bit about the communication protocol they use between their RC and Drone? Its not a standard protocol and not much info is available on the internet
Germany's had aced the technology at least 5 years before a DJI drone even existed. Their speed controllers were twice as fast and their platforms were super stable
Youโre telling me prior to 2011 no one had tried a PID controller for helicopters? I am in disbelief
Fascinating but too many undefined TLAs for this amateur to understand fully.
Quote
Christopher Nguyen โฝ
@pentagoniac
Replying to @kevinsxu
Frank Wang was an undergrad at @HKUST EEE/ECE when I was a prof there, and did his MPhil (and co-founded DJI out of HKUST) with my colleague Zexiang Li (also prof at HKUST) who did control/robotics under Lotfi Zadeh at UC Berkeley.
Ping Ko, then Dean of Engineering and another
Show more
I visited China for the first time in 2001, and this growth in creativity and innovation is probably the biggest change I've seen in China.
This is really cool but it does make me even more nervous about riding in a helicopter lol. Very cool story though, Iโm going to watch the vid later.
Why canโt we do the same thing with helicopters? Monitoring, adjusting, perfecting, etc?
Itโs amazing how drones are redefining so many things including surveillance, drug delivery and even farming
The breakthru was not mathematical, it was solid state gyroscopes and accelerometers, on a chip. These came out for commercial use with Wii hand controllers, and early Iphones. That is what made cheap drones mass market. I would add very cheap electo-optical sensors to that
Show more
a friend of mine studied aerospace told me propulsion theory was one the easier subjects while helicopter theory was the hardest
Nice to read on this but they had good Govt support to take this forward.
Et bien, je connais un ingรฉnieur chez Parrot qui doit halluciner s'il lit cet article
Shhhh, you gonna terrorise those westoids that believe technology is magic that can appear from nowhere.
Reminder that the cc3d devs had a falling out when the only dev that was actually making code was stonewalled (and then banned) when he wanted to migrate to 32 bit controllers. They were absolutely bought out by DJI.
I take it you have never heard of Ardupilot / Arducopter which started in 2007 with fusion of accelerometers and GPS to create a stable flight control system rivaling a full INS.
Your Wang was quite late in the field and didn't define the field.
A bold and provocative look at the hidden ideology behind AI and how businesses can harness its power without losing their humanity, from Chris Hood, author of the bestselling framework Customer Transformation.
amazon.com/Infailible-Art
It's curious why no one requests the link to Wang's 2011 thesis. Here is the link: lbezone.hkust.edu.hk/pdfviewer/web/
Discover more
Sourced from across X
![](data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" width="73" height="73"><rect fill-opacity="0"/></svg>)
GPT-4.5 is ready!
good news: it is the first model that feels like talking to a thoughtful person to me. i have had several moments where i've sat back in my chair and been astonished at getting actually good advice from an AI.
bad news: it is a giant, expensive model. we
Show more
ok fine maybe we'll do a social app
Quote
CNBC
@CNBC
Meta plans to release standalone Meta AI app in effort to compete with OpenAI's ChatGPT cnbc.com/2025/02/27/met
> Google: Released Gemini 2.0, gave access to everyone via AI Studio.
> Anthropic: Launched Claude 3.7 Sonnet, gave access to everyone.
> xAI: Launched Grok-3, gave access to everyone.
> OpenAI: New GPT-4.5 behind a $200 paywall.
Guess who has "Open" in their name? 