[Last night I went to the Silicon Valley Automotive Open Source meetup where Chris Anderson of DIYDrones/3DRobotics told his story.  Here are my notes, slightly rewritten and re-ordered to make more sense if you weren’t there]

DIY Drones talk by Chris Anderson, 3D Robotics

1/16/2013, Nokia Sunnyvale

Backstory involving building a Lego Mindstorms autopilot with his kids in 2008 (this is a good story worth watching, e.g. from his Maker Faire talk)

DIYDrones Community

• He fostered the community by “being dumb in public”, and letting interested people come and teach him
• DIYDrones is now the biggest community of people building drones, ~30K registered members, bigger than any aerospace company
• Ground control software written by an IT guy in City Hall in the Outback
• Smartphone processors and sensors have revolutionized robotics sensors and components
• Powerful, light, small sensors and processors are the “Peace divided of the smart phone wars”

Arduino

• people laughed at for the choice because it wasn’t the most powerful or professional
• chosen because it was the most open community
• lowered the barrier to entry – some top contributors are not even programmers or engineers
• got people that wouldn’t have participated if they needed an advanced tool-chain
• assumed the hardware would catch up
• Drone community should be like a great video game – easy to pick up, hard to master

Current models

• APM2.5 – universal autopilot for many helicopter and plane formats, GPS guided plus other sensors
• PX4 (ARM based) – enables cutting edge robotics research but easy enough for kids
• “We love the research dudes, but we really love the kids”

Military grade technology at toy prices

• uses: personal, classroom, disposable, delivery
• example – sending drones with biodegradable bodies on one-way missions to measure glacial pools
• feature parity with military (probably exceeded military features by now, plus better ground station), but sacrifice reliability to shed costs (currently ~2 nines for DIYDrones vs 8-9 9s for military, at .01% of cost)

Vehicle configurations

• Favor solid state components and minimize mechanical components
• Quad/hex/octocopters only have one variable: rotation speed per prop
• Solid state copter turning requires 200-400Hz adjustments in rotation speed, mechanical components can’t do much higher than 50Hz
• Mechanical parts are less reliable and more complex

Flight features

• Geofencing – the plane becomes uncrashable
• Mobile/tablet interface – touch interface, direct, focus on the mission
• “FollowMe Box” – Raspberry Pi radio tracker – turn it on, the drone flies to you, follows ~30′ above and behind you, filming, then when the battery is low it flies back to where it started. Get great windsurfind shots!

Free/paid

• DIYDrones non-profit community
• 3DRobotics for profit manufacturer

Manufacturing and production

• design and manufacturing in San Diego & Tijuana
• Tijuana is the Shenzhen of North America
• Electronics manufacturing – you need local production to innovate faster, iterate faster, don’t tie up capital in supply chain, don’t wait to sell out big batches, not hamstrung by errors

Contributor reward structure (pic)

Open hardware business model

• giveaway the bits, sell the atoms
• charge 2.6 BOM
• keep ahead of closers by innovating faster, supporting better
• 90% of features, 0.1% price compared to military
• democratize technology low price -> volume -> business

Open software hardware

• free fast better R&Dexempt from regulatory barriers
• development and support done by customers
• avoid FCC approval by selling components to developer
• regulations have exceptions for R&D and private use

More children flying drones than military service

• more sold per quarter than entire military
• ~30,000 ArduPilot modules sold vs ~7k drones in US military
• FAA regs: must remain under 400ft altitude, stay within line of site, away from built areas
• at that altitude, you get sub-cm photo resolution

Open source hardware challenges

• requires architecture of participation (hard to setup)
• untested legal questions
• reliability is hard b/c of complex hardware/software
• limited margins (cloners undercut pricing power)
• immature collaboration tools (version control for hardware/design)
• piracy

Amateurs can disrupt the biggest industries in the world

Q&A

What about suicide bombers – there are much easier ways to deliver bombs already

What about smuggling drugs? “Delivering drugs across the border is a solved problem”

Potential uses outside of recreation? Farms are a great place to test commercial uses, hiking search and rescue

What are the main constraints? Simplicity – they can do anything, but right now you have to enter all the vehicle constraints into the autopilot, there’s no onboard machine learning to teach the autopilot by flying

What about privacy? Privacy is a technological outcome that has been shrinking. There’s no national privacy policy, communities don’t control their airspace

Opportunities for involvement

• C++ for the hard stuff
• C/Arduino for control/navigation
• python scripting for ground station
• swarming – hard to program, but the hardware is there for drone to drone communication – all Arduinos are Ross nodes (?)
• new raspberry Pi station will have ruby/html5 interface (coming Feb?)

Ground station

• logs data
• record/replay flight info
• HuD on virtual horizon or superimposed over video
• map
• change input from radio, joystick, script, etc

Keeping things simple

• don’t tell the machine things they should be able to figure out on their own
• be simple but expandable; defining hardware is complex, support is hard
• they’re like PCs with driver problems, not an integrated Apple product, but that’s the price of being open
• maneuverability with as little mechanical as possible