Time to read: 5 min<\/p>\n
Successful product development involves a rigorous and detailed process of testing, breaking, and fixing. However without the right system in place, it can be difficult to parse out the details of effective changes versus ineffective ones.<\/p>\n
Nathan Hall-Snyder, VP of Engineering at Kespry, a commercial drone technology company that recently raised $10M in Series A funding, knows this process well. In the beginning of his journey with Kespry, it simply began with a curiosity around the drone space and then by a process of experimentation, exploration and problem solving, Nathan and his team have developed a robust system that allows them to rigorously test their products to avoid negative progress and inefficiencies.<\/p>\n
In this interview, Nathan shares his knowledge on how to implement a system for stronger, more reliable hardware products.<\/p>\n
Utilizing available hardware components and open source software is a great place to start for exploring the power of drone technology. Eventually, however, it has its limitations and the best way to develop a safe and reliable product is with strong control over both the software and hardware design.<\/p>\n
When asked about open source software for drones, Nathan is quick to express his enthusiasm for the DIY drone community and is open about the various tools Kespry used in the early days to explore what could be possible. \u201cI\u2019m certainly a proponent of DIY drone tools and in the beginning we experimented with a consumer drone just to help us learn and explore. We realized we could telnet into it to make it do things, which was pretty entertaining, and we learned a lot about takeoff and landing. Eventually we ran into some obvious computer limitations though, so going forward we realized we needed more computational power and moved to develop our own software.\u201d<\/p>\n
Because a commercial-grade drone requires a very high standard of safety and reliability, Nathan knew they had to move away from integrating outsourced components and toward a custom made drone. \u201cWhen integrating a number of subcomponents, like some other companies are doing, it\u2019s difficult to know what\u2019s going on when you run into a problem,\u201d Nathan explains. \u201cIs it a change you\u2019ve made or is it a problem with one of the subcomponents? It\u2019s hard to know in that situation where the problem originates\u2026 it\u2019s impossible to fix if it does in fact originate from someone else\u2019s component.\u201d<\/p>\n
Kespry owns their entire stack, including custom-built electronics, hardware components, and internal software. Nathan advises, \u201cWhen you own the entire stack the product is so much more reliable. We can test our entire system and control each change, which allows us to more easily and efficiently assess problems and integrate solutions.\u201d<\/p>\n
In addition to the necessity of custom-built components, Nathan stresses the importance of a detailed testing infrastructure. After working at Google X on their Google Glass product, he recognized the value of their rigorous testing system and wanted to integrate it immediately at Kespry. \u201cThe first priority when we set up our company was to find a system with which to integrate regression testing,\u201d says Nathan. \u201cThis way, any time a software change is made, it negates the possibility of making the drone fly worse than before.\u201d<\/p>\n
\u201cWe set up our testing infrastructure before we had even written any of our code for the drone,\u201d Nathan adds. \u201cThat\u2019s certainly a very Google way of doing things \u2013 to write the tests before you\u2019ve even touched the code! This way you can very easily measure your progress.\u201d<\/p>\n
In order to test the hardware, Nathan also uses a large test suite including a full drone simulator. \u201cWe fly the drone in a virtual space while testing all the sensor inputs, software, and hardware. We use this simulator to work our way up to a full hardware test,\u201d says Nathan. In order to test the hardware, Nathan also uses a large test suite including a full drone simulator. \u201cWe fly the drone in a virtual space while testing all the sensor inputs, software, and hardware. We use this simulator to work our way up to a full hardware test,\u201d says Nathan.<\/p>\n
A cautious and detailed testing process is surely a time investment, but it pays dividends down the road, saving you from the headache of unidentifiable problems. \u201cIt\u2019s all in proper testing.\u201d Nathan says. \u201cWithout it, you risk implementing one change that makes it worse and one that makes it better, leaving you unable to tell if you\u2019ve made any progress at all.\u201d<\/p>\n
Moreover, when reliant on someone else to assemble your parts, someone who doesn\u2019t know your product as well as you do, mayhem can ensue. \u201cThere can be so many problems when you set up parts and then have someone else assemble them,\u201d Nathan sighs. \u201cWhen I\u2019ve done that sometimes the product returns to me in a configuration I couldn\u2019t have even imagined! I\u2019m like, \u2018how did you even come up with that!?\u2019\u201d<\/p>\n
When it comes to building your product, Nathan encourages a hands-on approach, at least in the early stages. \u201cEvery drone I\u2019ve built so far I\u2019ve been doing by hand. Eventually we\u2019ll be ramping up our manufacturing process, but I really enjoy building all the drones myself because you learn a lot by assembling your own product,\u201d Nathan shares.<\/p>\n
At the end of the day it all comes down to knowing your product. The more you know about your product and all its features and parts, the better your product will be. Assembling a product yourself leads to a deeper understanding and is part of the process of more effective and efficient product iteration. \u201cWe\u2019re learning very quickly but there\u2019s always more to learn. We\u2019re just trying to learn as much as we can, as quickly as we can, to build the best product possible,\u201d says Nathan.<\/p>\n
Develop your own components, build and assemble them yourself and test them continuously and rigorously. It will be a detailed process, but you will thank yourself at the end for the product that results.<\/p>\n