Of all the robotics megatrends ubiquitous connectivity is the one that arrived fastest and that is already having a major impact in deployed robotics. Constant contact with their robots allows companies to ship the robots earlier in the design process and iterate faster than they could before.
For any deployed system, uptime is one of the key performance measurements. Online services companies will often brag about the “Five 9’s” of uptime, meaning that they are online 99.999% of the year (or down about five minutes a year). In any complex system, the last 2% of uptime takes an enormous amount of effort. Demos are easy, but production quality is hard. With connectivity, you can deploy robots when they are only 98% reliable, as long as that last 2% involves problems the robot can detect and it can fail safely.
In general, anything that can save time and money makes it even easier for startups to take on these types of capital-intensive ventures. The three core areas where connectivity can assist and transform robotics startups are: (1.) real-time data collection, (2.) human-in-the-loop control, and (3.) the ability to update robots in the field.
The first time we really got a glimpse of high-speed cell signals was the roll out of 3G networks less than a decade ago. (The iPhone 3GS came out in June of 2009.) Even then, those networks weren’t broadly deployed, coverage was spotty, and high-speed was a bit of an aspirational term. Today we’re living in an LTE world, in which nearly the entire country is covered in LTE signals. To top it off 5G is starting to be deployed and that standard will be able to support up to one-gigabit speeds.
Why does this all matter? Previously when you deployed systems, the first time you knew something was wrong was when the robot didn’t return or an angry customer called to report a failure. Constant connectivity gives a company the ability to do preemptive maintenance, improving overall system uptime and lowering costs. In addition, for any robot that will need a large data set of training data for machine learning (such as mapping or object identification), thanks to connectivity, the company can compile and train on that data in real time.
The second major area where connectivity can be transformational is in human-in-the-loop control. This won’t be seen in real-time teleoperation, as the latency is far too high to control in real time. What you will see is something similar to what NASA’s Jet Propulsion Laboratory has been doing with Mars Exploration Rovers for a decade. The remote operator sets waypoints and objects of interest, but the robot itself determines what to do in real time for object avoidance and path planning. This means that a company can deploy a mobile system while they are still fine-tuning more advanced algorithms, such as object classification or prioritization in a dynamic environment.
The last area where connectivity is a game changer is the ability to update a robot in real time. Tesla has shown that a company can push safety-critical updates in real time as soon as they discover a bug. It’s hard to overstate how much this changes a company’s deployment strategy. Previously, people would have to overnight courier hard drives or put people on planes to be able to update robots. And the better the company was doing in terms of robots in the field, the worse the problem was. Now companies can collect incident data in real time, solve the problem, and then push an update to all the robots in the fleet.
The best part about ubiquitous connectivity is that it is a megatrend that requires zero investment from the robotics community, thanks to the telecommunications industry’s 5G roll out.
I’m very interested in finding more hybrid systems that are using a human in-the-loop to help control many robots at the same time, scaling up to systems where one human is controlling up to 100 robots at once. Our portfolio companies like Simbe and Marble have examples of robots that can leverage ubiquitous connectivity to deploy earlier and still maintain a high reliability in the field.
If you are working on something like that or robotics in general, please feel free to reach out to me at email@example.com.