Ease of Programming
The second challenge is how difficult it is to actually create products around these platforms. Imagine how many people can program a mobile phone today. It’s so easy school-aged children can make great apps. Now imagine what it was like when people had to program computers using punch cards.
It’s an exaggeration, but robotics today is the equivalent to that. Not only do you require a team that understands electronics, circuits and wiring – you also need software developers that understand firmware, people who are experts in machine vision, machine learning, security, wireless networking- the list goes on – but essentially a team of PhDs.
All of this effort gets people to the point where they have a functional robot to start building on. People consider much of this re-invention their IP – even though every other company builds a specialized version of it themselves. The software that people are trying to build to help with this is too decoupled from the hardware – for now. Despite open-source libraries like ROS, OpenCV, Eigen and others, there’s still lots of extra work required to get an actual robot working.
Connectivity, Data Are Afterthoughts
The third challenge is that wireless connectivity is always a bolt-on addition. An afterthought. People just think they can stick Wi-Fi on these devices and everything will just work. That’s how we all access the Internet with our laptops, right?
Well, what happens when the robot moves outside? What happens when two robots connected together move farther than a couple hundred feet apart? So they change their mind and bolt-on an LTE modem. Data plans are expensive. For connected machines, cell-phone companies often only provide around 1Mb per month for a reasonable price.
For that price your connected robot might be able to send a few crappy images per month to other robots, or back to you. This isn’t very useful for robots that need to make quick decisions in tandem with each other, or if you’re a company using robots collecting data to make decisions (for example natural resource discovery).
The biggest challenge of all is interoperability. None of the products you see today, or in the near future, are built to work with one another. None of the drones you buy are built to easily work with your automated lawn mower, or your autonomous car, or your fridge. The list goes on.
You might be skeptical that these things need to work together. There certainly may be some advantage to it – for example, an idle autonomous car could pick up your groceries. But it’s more about the ability that they can work together. If they can talk to each other, and fundamentally work in the same way, the creators can spend less time just getting up to speed and more time working on the hard problems.
Imagine if you had to design a circuit board, program the operating system, touch screen, wireless connectivity and case just to build an app for a phone – that is the current state of robotics today.
Jason Ernst, PhD Candidate, CS, is the CTO of Redtree Robotics. Ernst took first place in CODE 2014, Canada’s largest hackathon with 930 participants.
Redtree Robotics develops the Hydra chipset, which is designed for communication between more than one robot. Groups of robots powered by the Hydra chipset are always connected and can easily share data.