Haddington Dynamics
Dexter Haddington Dynamics
Haddington's Dexter robotic arm.
Two questions are parried in quick succession by two humans sat in a back room in formation with several robotic arms. Todd Enerson and Kent Gilson, the President and Inventor at Haddington Dynamics, respectively, are front and centre, disdainfully citing the NDAs adorned with their signatures constraining their ability to venture into the details when asked ‘what does company x do with your Dexter robotic arms?’
Haddington Dynamics counts hospitals and medical centres among its customer base, automotive and aerospace giants too, and even GoogleX housed one before the employee who purchased it went off and launched his own robot company. But what the Dexter robotic arms are being used for by these organisations is off limits, they’re picking and placing, used for imaging and delivering, but exactly what is apparently, and regrettably, none of my business.
A few years ago, there was no business at all. Gilson founded Haddington Dynamics in 2015 but not before years were spent trying laser cutters and desktop extrusion printers to build a robot. To little avail. Gilson had one of the first original Replicator machines to come from MakerBot’s Kickstarter campaign. He had also attempted to print parts on a FlashForge, before finally finding joy with a Prusa machine.
“We actually got good value out of the Prusas but the other ones, it was nightmarish, we didn’t have a business,” Gilson recalled. “The difference in product tolerances and part tolerances were different between the beginning of the roll and the end of the roll, let alone from machine to machine. We had to spend an enormous amount of time recreating the geometry.”
Yet, it was time the Haddington team was willing to give up. They were committed to the idea of building algorithms to adapt the geometry in accordance with the limitations of 3D printing, but during a customer ‘makecation’ at Haddington’s Las Vegas office, a NASA employee brought with them some parts printed on a Markforged system. 3D printing was improving, and by April 2019, the company had purchased 11 machines from the Boston-based vendor, two Twos and nine Onyxes, with the goal of getting up to 40 by the end of the year.
“We’re seeing as good a part as you can get on a milling machine and then what really took us over the edge was the embedded carbon fibre, the continuous carbon fibre, and we knew that 3D printing was going to reach a threshold where the things you could build with it would outperform fabricating something in any other way,” Gilson said.
About 18 months ago, Dexter, which boasts 3kg payload capacity, a 1.4m spherical work envelope, and 0.03mm load adaptive repeatability, was completely redesigned over a two-week period per the capabilities of the Markforged machines. A total of 800 parts was consolidated and reduced to about 183, of which around 85 are 3D printed, grippers included. The components that aren’t printed include bearings, nuts, and bolts, but there’s now a goal to implement these parts, as well as the electronics, as builds are taking place. Taking Dexter’s dual spate cycloidal drive as an example, the aim is to pause the print as and when necessary to insert bearings, nuts, and bolts, before resuming the print so when the component comes off the build plate it’s a fully functioning transmission. Making those insertions will be other Dexter robots.
Haddington Dynamics
Dexter Haddington Dunamics
Inventor Kent Gilson with the Dexter robotic arm.
Haddington is looking to promote collaborative workspaces where humans work in tandem with robots to create more robots. As of April, there were 15 humans, God knows how many robots, and nine 3D printers running around the clock. It’s a process by which robots take care of the menial tasks, humans handle the more thought-driven jobs, and the printing systems facilitate the reduction in the number of parts, tackle complex designs and save weight.
“Our whole process is organised around using the Markforged as a production machine,” Gilson explained. “When you build robots, power to weight ratio is one of the four corners of optimisation that you’re trying to balance. If you’re going to go with injection moulding, then you’re going to have to design for injection moulding, [but] you don’t get the advantages of the free complexity.”
Printers run overnight with parts scheduled to be completed for 8am so when the team punches in, the beds can be cleaned, parts checked into inventory (bar codes are implemented for assembly), and more parts can be produced. Around 80% of the printed parts can be produced on the Onyx machines only, with the Twos helping out on the remaining 20%. Dexters can be manufactured in just over a week, the unskinned version selling for $6,800; unskinned with a gripper for $7,400; and fully skinned with a gripper for $9,800. Such is the price of these products against the expenses involved in purchasing its 3D printing systems of choice, Haddington reckons it could expand its capital equipment by around 150 a year, and still amortise the cost.
The company also offers a series of end effectors for between $250 and $850, circuit board packages for $895, a Makecation for $2,000, and everything you need to build a Dexter HD robot yourself for $4,800. Haddington doesn’t want to just make robots but make robot makers. And the enabler of that objective is another Haddington quirk.
“The entire robot is open sourced,” explained Gilson, “including the FPGA code that implements the control system, all the geometry, all the drivers, all the Dexter development environment which is a java script ID based software tool.”
Haddington decided to open source everything around its flagship product after winning the Hackaday Prize last autumn, exposing the company to some honest assessments from the community, and sceptical questioning from venture capitals. The latter is being batted away, with the point being made that ‘you can’t really steal it if we open source it,’ and ‘we also have first mover advantage.’ What’s more, if someone did begin to make Dexter robots cheaper than Haddington can, the company says it would simply turn into the customer and apply the robots themselves. Provisions have also been put in place in the form of a GPL 3 license so if any customer wishes to alter the design, it has to either remain open source or pay Haddington for a close sourced commercial license.
Haddington Dynamics
Dexter Haddington Dyncamis
The company is committed to community, transparent with its process and expanding its network. A dialogue has been opened for questions to be answered, and answers to be questioned. Users are encouraged to share experience on Haddington’s online forum, good or bad, to help move the project along.
“We’re happy with fail mechanisms, we’re happy with failure mechanisms,” Gilson said. “Our claim to fame is we f*** up faster than anybody we know. You need to be able to handle that and keep your pride out of it and be thankful for people finding things that we can’t afford to hire people to find.”
The result is a mutually beneficial relationship between consumer and manufacturer, one that could flip on its head should another team hit on something more economically efficient. Haddington is settling into this open source environment, where the technology is as wide open to access as it is to critique. It is also thinking about its future product development, eyeing the proficiencies of generative design software to optimise designs for specific tasks, and do so automatically. It wants to automate the whole line of production, have robots making robots, picking and placing repeatedly so its humans don’t have to, for example.
Its vision is not just to keep this in-house and make its own manufacturing set up more advanced and seamless, but to encourage its community to do the same, giving them the tools they need, the direct access of communication, the means to become robot makers who oversee robots making more robots.
“We are trying to create a friction-free adoption mechanism,” Gilson said. “We wanted to make enough profit to encourage other people to actually start building robots also. That’s our goal, we’ll build robots until other people come up to speed and can build robots. Our vision is to get 10,000 individuals or small companies out there building a few dozen robots a month to supply the community with these motion and force platforms.
“It’s really about taking advantage of this trend towards Industry 4.0 where manufacturing is going to move to the edges as opposed to centralised and spoken hub distribution on top of massive single item factories. We’ll ride that, we’ll provide robots on the front end of that and also, we know as the robots start manufacturing themselves and our robots are able to move more precisely than the parts they’re made of we’re going to be able to fabricate our own parts and assemble them. That’s going to drive the commoditisation of this type of product faster than any other type of product, so our customers are going to be these people or businesses that we create that are actually building robots.
“We want to become a central supplier of parts and know-how and social capital to drive this into these thousands of factories. We do that by providing them with the technology that’s got market value and profit potential and then supplying them with the means of production to build that, supplying them with the software that turns their robots into more efficient means of production to make more robots. It’s a strategic plan to get to a much bigger business than just selling robots.”
