ArduCopter Hexa with camera mount, by jDrones and Chris Anderson.

Anderson sees a future where robotic aviation is ubiquitous. He created the first DIY kits using the ArduPilot, then founded 3D Robotics, a company making parts, kits, and ready-to-fly UAVs. We asked him what he’s learned.

You’re a DIY robot enthusiast who became a hardware manufacturer and kit maker. Any lessons to share with aspiring kit makers?

Amateur assembly: Chris Anderson's kids.

The biggest lesson is that I should not be creating kits. The DIY Drones community is developing amazing technologies like the ArduPilot. We’ve created a business with several large factories creating kits, and they do it much better than me and my kids.

I’m chair of 3D Robotics, which makes the ArduCopter kit brain, and we work with others who make their own kits. We’ve got more than 120
different SKUs. It’s a multimillion-dollar operation now; some are kits, some are just boards.

What tipped your decision to sell UAV kits, rather than just share designs?

Three or four years ago I built the Blimpduino. I had the idea that you could have an autonomous blimp that would fly around in schools, and it would be great for teaching kids about robotics.

So we designed the board, and we put the Eagle files up, and said here’s a link to SparkFun’s BatchPCB, here’s the parts list to buy from Mouser and Digi-Key, and here are the instructions for putting it all together.

That’s when we realized it was too hard — people just weren’t going to do it. We had to build the board for people.

I made about three boards and said, “I’m never doing that again.” So we got a contract manufacturer, and found out that you really have to order at scale to get pricing. We got the money together to buy 500 boards — you have to make a leap of faith, take some capital risk to get the volume. It was a little scary.

Then we realized that sourcing the other parts is also hard. Regular folks have no experience in it. I spent a ridiculous amount of time learning the economics of mylar balloons. The learning curve is steep and expensive.

Kitting it was the only way to really get it into people’s hands.

What did it take to make that first kit?

An ArduPilot Mega board is born.

Eventually we got all the parts, got the boards built and tested, the firmware loaded, labels printed, and at that point I bribed the children for $1 an hour. We labeled all the parts and set up an assembly line on the dining room table. I was doing quality control, but in the end, despite all our efforts, about half the kits were missing a part. I spent months sending out extra Lego parts or motors to people. I think it’s cute that a 5-year-old forgot to put in a part, but the customer might not be as charmed.

Now we have factories in San Diego, Tijuana, and Bangkok, running professional assembly and packing operations so we can keep the customer happy and actually keep these things in stock.

What lessons did you learn the hard way?

Sourcing parts is expensive. You have to buy wholesale to sell retail, which means buying in volume — in the thousands to get decent pricing — and that’s capital risk, especially if you make a design mistake. You have to make a bet that you’re going to sell a lot of them. You have to explain to your wife or husband why you just put $10,000 on the credit card.

Hand assembly means you’re in the assembly line business, and you’ll spend a lot of time making sure everything gets into the kit.

Often, worse than failing is succeeding. This is not a one-time exercise — if you create 1,000 kits and succeed, you’ve got to do it again! The third time, I had to pay the kids way more than $1 an hour. As a kit maker you may discover that the worst thing that can happen is that these things become popular: “Oh no, we can’t keep them in stock!”

Don’t forget to sell it for a profit.

What would you do differently?

Pro assembly: Arturo in San Diego factory.

If you really want to make a kit that lots of people will enjoy and you’ll continue to sell over the years, it will quickly get too big for the kitchen table. Find a commercial partner that really cares about quality control. It’s a real business involving paid professionals doing quality assurance, answering customer support calls, ensuring that this thing is good and stays in stock. As enticing as it is to create a kit, it quickly becomes unfun if it’s successful.

Also, the tech support on kits is a nightmare. It imposes a huge aftermarket burden: you’ve got to help people fix it, or take back kits.

In this issue, MIT’s Michael Schrage says that kit makers like DIY Drones suggest a robotics future “more varied and ‘out of control’ than anything envisioned inside the Pentagon.”

Do you see amateur UAVs getting out of control? (I’m thinking of the so-called drone terrorist arrested by the FBI.)

First off, there’s no evidence that that guy had GPS or anything other than an R/C plane.

An assembled DIY Drone with parts totaling about $300.

Openness means most people will use it for good, and some will use it for bad, but that’s true of any tool: computers, cellphones, even hammers. There are those who want to ban technology and ban the maker instinct.

So what’s the future of UAVs in the hands of amateurs?

We started with kits, but we’re very quickly moving toward Plug and Play — there’s two orders of magnitude difference in the sizes of those
markets. If you have to put two parts together, you’ve eliminated half your market — and if you have to solder something, you’ve eliminated 99.9 percent of your potential market!

-Keith Hammond

To get Herbie to work, we did have to change a few things from the directions. They tell you to use double-sided foam tape (provided) to attach the motors. However, when we did that, we couldn’t fit the battery in. We just removed the tape, reheated the solder to move the motor closer to the board, and glued it on with Sumo Glue. We also discovered that Herbie will not follow LED flashlights; they just don’t put out enough infrared.

The Boe consists of an aluminum frame, upon which two servomotors, a battery holder, and a BASIC Stamp 2 microcontroller board are mounted.

Over the years Parallax has developed numerous add-ons. One I like best is the servo-operated gripper, which makes the Boe-Bot look like a nasty pincher bug. The gripper can lift small objects like chess pieces and ping pong balls, but it won’t fetch a beer — unless it’s a very small beer from a very short fridge.

Suitable for first-time builders, the Boe-Bot takes about 2 hours to build. The gripper kit requires a bit more skill; plan on another hour. Other nifty accessories include a six-legged walker kit and an infrared line-following module.

This entertaining kit, suitable for kids 12 and up, lets you build three different tracked robots. You start by assembling separate modules, then snap the modules together to create a rover, forklift bot, or gripper crawler. You can always unsnap the modules and reconfigure them for another style. I built the forklift version, spending about three hours while watching reruns of Star Trek. The assembly manual is clear and concise, with great 3D pictures to show how things go together.

The robots are motorized using a wired control box. Two motors operate the left- and right-side tracks, while a third motor operates an extended mechanism — the forklift goes up and down, for example, and the gripper opens and closes.

The ATR is immensely hackable, thanks to its modular design. Unplug the 4-pin connector from the control box, and substitute your own microcontroller. Looks like there’s just enough room on top for an Arduino, a motor shield, and some LiPoly batteries. Seems like I’ll be watching more Trek episodes!

Reviewer Tyler Bennett of Cary, N.C., writes: “This kit was a lot of fun for me as well as my 11- and 12-year-old grandkids. The kids put the kit together (with me looking over their shoulders) in around two hours and then we all enjoyed running it around on the floor.”

Most useful of all, a remote control menu offers direction and speed control for two motors and an action button to control one motor independently. This lets you test motors without having to program anything!

I’d like to see more programmable storage memory, but overall this is a nice refresh for Mindstorms NXT. It’s an excellent kit for budding robot enthusiasts and also a great gift for veterans.

This is not a project for beginners, but thanks to clear documentation and well-organized components, anyone with moderate soldering skills and sufficient patience shouldn’t have much trouble. It ships with analog “discrete brain” circuitry, but “brainboard” kits let you add popular microcontrollers.

Topobo is part construction toy, part kinetic memory robot. By combining solid parts with active motor/brain hubs (learning servos) you can spend hours creating unlikely creatures with even more unlikely forms of locomotion.

My young kids immediately got how to record and play back the motion of their creations; press a button, perform a few poses, press the button again, and it’s alive! I wanted to save movements between sessions, but my kids understood before I did that Topobo is about exploring novel methods of movement, not reproducing past creations.

Our only complaints: the Lego Technic-compatible connectors are hard to remove, and we’d prefer batteries to being tethered to the wall with the power supply wire.

All arms in the product line are based on the company’s Servo Erector Set parts. These are prefab aluminum pieces designed to mate with standard R/C servomotors. You can buy SES parts separately, too, for your own arm designs, or to modify one of the stock arm models.

The AL5B is for experienced builders. The hardware comes in numerous bags, and it can be a trick keeping track of everything. I suggest you open each bag only when you need it, and keep a measuring tape nearby to check the length of screws. Study the online assembly instructions ahead of time while you wait for your kit to arrive. The kit doesn’t require soldering, but needs a base you supply. I used a piece of ½”-thick plastic.

Included in the AL5B kit is Lynxmotion’s SSC-32 servo controller board. Connect it to your PC (via serial cable or USB-to-serial adapter), and run the free FlowArm software to control the motion of each joint. Record
complex motions and play them back with the click of a button.