(Following is an unedited webinar, presented by Robert Dahlstrom is founder and CEO of Apellix Robotics.)
Today we’re talking about robots, and more specifically software controlled robots and how robots are software.
As we know, robots and certain drones are definitely hardware, but what does hardware require to run and operate? Obviously, the answer is software.
Robots are also machines, as well as hardware. According to the Oxford Dictionary, a machine is an apparatus using or applying mechanical power, and having several parts each with a definite function and together performing a particular task.
But in addition to being machines, robots are programmable machines. Wikipedia defines a robot as a machine, especially one programmable by a computer capable of carrying out a complex series of actions automatically. It’s interesting in this Wikipedia definition that they specifically say a machine programmable by a computer.
These robotic systems are machines, programmable machines, and automated machines. What a machine does is, a machine uses power to apply force, or to control a movement, or to perform an intended action. They can be driven by anything from animals, to people, to wind, or chemical, or thermal, or most commonly electrical power.
These machines include system of mechanisms that takes the actuator input to achieve a specific application or output of forces and/or movement. These include computer sensors quite often that monitor the performance of these machines, that are able to plan movement in these various mechanical systems.
Way back in the beginning of the renaissance in Italy, philosophers defined six simple machines, which were the elemental devices that put a load into motion, calculated the ratio of output force to input force, and these are known today as mechanical advantage.
You can think of something as simple as a lever, and the leverage you can apply from that, so that mechanical advantage put in to an automated hardware machine is what enables us to come up with the robotic systems that we have today. These modern machines are complex systems that consist of structural elements, and mechanisms, control components, interfaces for convenient use, anything from automobiles and airplanes, those are machines, to appliances in our house as well as farm machinery, and more along the robotics area, factory automated systems.
Most of these robotic systems utilize microprocessors. Outside of the small subset of mechanical only robotic systems, the microprocessor is a core component of robotic systems.
Microprocessors run by having embedded software built into them. Without that software, a microprocessor control program will not work. What that enables is, it allows for use of performances unattainable by humans.
One way to think of this is mechanically calculated mathematical profs, such as solving the equation for pi. How arduous that is as a human, and how simple that is for a processor controlled machine like a calculator and/or computer spreadsheet.
This embedded software that’s in these machines and mechanical devices allow us a lot of flexibility in what we can do. By embedding a software controlled microprocessor, it allows to tailor for different needs of a product line. You could upgrade to performance with minimal redesign to the product. What I mean by that is making modifications to the software, to the embedded programs within these microprocessors actually enables additional functionality that is independent of modifications to the hardware of the machine itself.
The microprocessor control systems also provide control strategies that are impractical to implement with electromechanical controls, or with purpose-built electric controls.
Examples of this would include motor-control systems where you can adjust based on timing, or speed, or load, or ambient temperature, or other variables.
This comes back to my core argument here that robots are computers. Because they’re using microprocessors and computers are microprocessors. A microprocessor is defined as a computer processor which incorporates the functions of a computer’s central processing unit, commonly called of CPU, on a single integrated circuit, or multiple integrated circuits.
Given how most robots, most modern-day current robots have multiple processors or have, in addition, have multiple sensors, and those sensors contain processors, therefore, I’m going to say that robots are computers because when you remove the computer aspect form a robotic system you’re left with a not modern robotic system, you’re left with something in essence a mechanical design.
Microprocessors are programmed with software as we know and they don’t work without it. Since robots contain these multiple processors, they’re computers. Here is a picture of a robot called Sophia, she’s a humanoid robot that uses artificial intelligence, developed by Hanson Robotics, and she’s designed to respond the questions and it’s been interviewed from reporters from around the world. The reason she’s notable is just in October of this year, October 2017, she became the first robot to receive citizenship of any country, Saudi Arabia granted her citizenship.
Since robots are hardware, like we started out showing pictures of hardware, since robots are machines, we started out showing pictures of machines and talking about them, and since robots are programmed automated machines utilizing microprocessors, my argument is that robots are software, robots could not work without software.
What about some of the newer software technology such as machine language, visual processing, artificial intelligence, is that software? Absolutely. What about human intelligence, is that software? Some people say yes. The analogy is the human body is the hardware, but the brain is the software, which is the electrical-chemical neurons that focus on things.
It’s interesting to have this cognitive structure to look at things as a hardware being the physical part of something such as a robotic system, and software being the brains of something like a physical component such as a robot.
Now that I’ve been droning on for a bit, I just want to talk a little bit about how drones, what they are, and how certain category of drones are robots, and other drones are just mechanical devices.
There are several definitions of aerial drones, different mechanical, types of aerial robots, or mechanical systems. But what to me makes a drone a robot, and that is a drone that has a ability to do something other than just fly and take pictures. Now, drones, when I’m speaking about drones, this includes more than just the common perception of the drones that fly, but also the ground vehicles and underwater vehicles.
To me, the drones that fly and take pictures that are common in today’s world are not robotic, aerial robotic systems, to me a robot, a typical manufacturing robot, has what’s called an End Effector. However, if you have an aerial robot that uses a drone for the lift platform and you add technology from the robotics domain, such as an end effector, to me then you’re able to call that drone a aerial robot.
Robots are amazing, they can do all sorts of new things in addition to what they’ve been doing historically, including adding safety and life features. For example, there are robotic systems now that are medical, that do medical procedures that are difficult or dangerous for humans to do because we have additional precision with the robotic system that’s not available with the human doing it.
Similarly, with the aerial robotic systems, you have the ability with some drones to do things like, in combination with a ground-based robot for example, control a bulldozer, they actually are systems in use in the world right now where there’s a drone that’s flying in the air, unmanned, autonomous, there’s a bulldozer on the ground, unmanned, autonomous, the two are communicating with each other, and completing tasks.
It’s very interesting how we were able to get to this place where we are in history. There’s been a lot of enabling technology that brought this around. Part of this has been fueled by the ubiquity of cellphones, and the massive economies of scale. Miniaturization of all the computing processing that goes into your cellphone, and the example also of Moore’s Law where computing processing power doubles approximately once every 18 months.
Lots will argue how we got here with software. Without the software that we have, modern software, we wouldn’t be where we are today. One quick example of that is the computer industry. If you’ve been around in the early ’90s, when personal computers were first being utilized in business, you notice that the software has become more complex, and there are those that argue that the software has actually driven the development of the hardware, pushing it to keep up with the capabilities.
The example being the visual operating systems like Microsoft, before that with the text-based displays on a computer, the hardware required to run that software was much simpler, one you bring in the graphical interface for the software, it required additional hardware requirements. Software, in part, drove some of the development of the computer hardware.
There’s a really interesting book that talks about this titled The Second Machine Age. The authors of this book argue that the second machine age involves the automation of a lot of cognitive tasks that make humans and software-driven machines substitutes rather than complements. We’re contrasting this with what’s called the first machine age commonly called the industrial revolution, which helped make labor and machines complementary. This relates to where robotics is now and where we’re going with robots and software development of robots in the future.
This rapid changes of the second machine age where automation of cognitive tasks are starting to occur is enabling a lot of technology that we in the past thought was 10 or 20 years in the future, right now it’s an exciting time of a lot of brilliant technologies.
Part of how we got there is exponential growth. Thinking back to Moore’s Law, and how computer processing doubles every 18 months, there’s an interesting fable called Rice and the Chessboard fable. The fable goes that the inventor of chess showed the game to the emperor of India. The emperor was so impressed that he invited the inventor to name his reward, he said “This is a wonderful game, I will grant you any boon you would like.” The inventor said “Oh, I’m just a humble game maker, I only wish for this, one grain of rice for the first square, two for the second square, four for the third square, and so on.” In essence that’s showing exponentially how things grow and double with each increment.
The emperor agreed thinking that this was a reasonable request. What happens is after 60 doublings you’ve got this huge number, actually, with 32 doublings, halfway through the chessboard, you have more rice than has ever been harvested in the history of the world. If it was stacked in a pile, it would be taller than mount Everest.
It’s kind of non-intuitive and hard for humans to understand how exponential growth works because we’re used to doubling things. One other example would be, if I was to take a step, I know how far that is, if I was to double that and take two steps, that’s also not that far, I know how far that is, if I was to double that, and take four steps, that’s not that far, but what happens is that doubling extends quickly, so that it would actually take 65 million steps to circumnavigate the globe, the world, which is only 27 doublings.
As the graph for exponential growth shows, this is just sort of typical for humans to wrap their heads around, because we don’t have experience with exponential in the physical world that often. As you see, this graph shows what is typically referred to as a hockey stick curve, and it’s the natural exponential function of y equals e x, e to the x power, where x and y are points on the graph.
Where are we now with Moore’s Law? We’re about halfway through the chessboard. Think of things that are very heavy processing microprocessor-based calculations and how everything that we have today will be able to double that same amount of processing in the world, this time next year, or in 18 months. That is leading to a lot of the changes we’ve been seeing with robotics systems.
This exponential growth of computing power has helped enable what’s been referred to as the fourth industrial revolution. The fourth industrial revolution is a collective term embracing a number of automation, data exchange, and manufacturing technologies that draw together what’s been called cyber-physical systems. This includes such as the internet of things, or the internet of services.
The industrial revolution, also called industry 4.0, is characterized as a digital revolution. It’s blurring the lines between physical, digital, and biological spheres. This internet of technologies example is promising to disrupt the global economy and create a lot of growth and prosperity, and there are a lot of businesses that want to participate in the fourth industrial revolution, that are trying to transform their businesses to get onboard by going digital.
Again, none of this works without software. Back in 2011, Marc Andreessen. Marc Andreessen is one of the founders of Netscape, one of the first browsers that was used on the internet, and he co-founded one of the largest general-partners of Silicon Valley venture capital firm Andreessen Horowitz. In 2011, he said “Software is eating the world.” This is a fantastic article to read because it lays out just what I’ve been talking about in this presentation that without the software power that we have, we wouldn’t have been able to advance as rapidly as we have, that we wouldn’t be able to continue to advance as quickly as we are able to now.
This was revisited in 2016 by TechCrunch and they said “Marc Andreessen penned his famous ‘why software is eating the world’ essay in the Wall Street Journal five years ago. Today, the idea that every company needs to become a software company is considered almost cliché.”
As software is eating the world, it’s quite amazing what all the things that software can do, and from a large part software is responsible for robotics, and how we’re able to do just amazing things with robotic systems. As the software increases, as the software becomes more sophisticated with artificial intelligence and neural networking, as the hardware becomes faster, more capable, as the communication between these things becomes more ubiquitous, and easier, and we have things such as swarms of flying drones, and warehouses these autonomous robotic systems that are navigating independently of one another, autonomous cars. All this is exciting, and it shows the value of these robotic systems and specifically the value of the software that enables these robotic systems to do what it is that they’re able to do.
I know I’m droning on about software, but it’s just to me incredibly exciting the things that you can do with software-enabled hardware.
The potential of robotic systems is absolutely incredible. The things that are coming up that we’ll be able to do will be things that we have not thought of in the past. Robots are a young and multifaceted industry, and they’re only now just beginning to show their true potential.
This brings me to what I’m calling the platform, so I can describe the importance of a platform. What do I mean by that? Let’s take the examples of iPhone, or mobile phone. The iPhone is a platform, meaning that you have hardware, and on that hardware you have a platform to run a multitude of software applications. You have a software application that allows phone calls, you have a software application that allows you to check your email, you have software applications, you have the App Store, where you can download all these things that make your phone do things.
When the designers of the iPhone were building it, they put it the camera, and they put in the flash on the camera. But because this is a platform, and because this is software, somebody came up with an idea that the designers of the hardware never would’ve thought of, and that is an app that uses the light on your camera to shine, if you put your finger over it, will shine through your skin and actually show you your blood pumping through your skin and give you your pulse.
That’s something that shows the value of software because that’s something that was never intended to be designed in by the hardware creators, but because there’s a platform that enables the software creation of this, it shows that you can create things that never were conceived of just by modifying software.
The platform is everything from you’ll hear the terms such as Software as a Service, or Platform as a Service, or Infrastructure as a Service, X as a Service, which is meaning everything is a service.
A lot of companies now use this platform, be it an iPhone, or robotic system, or mobile device, they use a platform to provide these various services.
Again, to me, this only happens because of the software. Remember, hardware doesn’t work without software. Software is pushing the boundaries of what hardware can do and incredibly making the hardware even better. Another example of this would be Tesla cars, the electric cars. They made a modification in the software for the algorithm downloaded it to the cars and increased the acceleration of the cars. Same thing with the right hardware on the cars, Tesla could download software and your car can become a self-driving car.
It’s actually interesting, I was talking about iPhones earlier, but in Silicon Valley, some of the venture capital firms even refer to them as software wrapped in plastic.
Workers’ safety, as I touched on earlier in the presentation, is a very critical and important part of robotic systems. Not only can robots work in conditions when humans can’t, where it’s maybe on a factory floor, where it’s hot, or there’s hot welds flying different places, or there’s gases and/or lack of oxygen, none of that really matters to a robotic system. Robotic systems are able to make the world safer for people.
As we continue using robotic systems to innovate towards safety, it’s interesting to think of how we’re able to bring that efficiency of a factory out into the world. That’s what we’re doing now with robotic systems. While robotic systems mainly started in factories for automating tasks on assembly lines, and they’ve gotten more sophisticated, as we continue this development path of faster processors, cheaper processors, better sensors, software that can make sense of all this and put it together in new and novel ways, we’re able to move these robotic systems from the factory out into the real world for real world applications.
The other huge advantage of this is that this enables us to replace human judgment with science, because these robotic systems with various sensors on there are able to make judgments based on scientific data rather than human anecdotal data, or information.
This revolutionary technology of software that came customized and enable robotics to do things they’ve never done before pertains for a very exciting future. Basically software rules. You can say that literally, because software is, in essence, a set of rules controlling something.
Imagine machine operator from the safety of ground operating robotic system that can do something, that’s another example of revolutionary technology, and where we’re moving with this ability of this new hardware, and this new software to do things that were undoable in the past.
A big example of this is the drone, like I spoke of earlier, drone with a End Effector on it is considered by me an aerial robotic system. We’re now able to put these aerial robotic systems manually do tasks that remove workers from harm’s way. One of the common causes of death in the US is falls from height. If we’re able to prevent that, if we’re able to put a robotic system in place to do something that a human is doing that’s dangerous, it just makes a lot of sense.
I was speaking to the CEO, who’s worked in industrial cleaning of oil, and gas, and chemical facilities in his career, and he said “Bob, I’d had to tell five mothers that their sons are never coming home again.” If we can prevent those conversations from happening by putting a robotic system in place, doing the job that a human used to do that was dangerous, then that’s a huge win.
Workers’ safety is incredibly important. There are companies, especially in oil and gas, infrastructure, and maritime, and other energy companies that actually start out every meaning with a safety briefing. It can be the business executives meeting in a conference room, but they start out with a safety briefing.
A lot of industrial companies are very concerned about safety. OSHA, the Occupational Safety and Health Administration of the United States has what’s called a hierarchy of fall protection given that falls are the number one cause of death and workplace injury in the United States. The OSHA hierarchy of fall protection, they say, starts with eliminating the hazards and risks of falling by engineering them out and away from the workplace. That to me is a perfect example of what modern robotics technology can do.
Software-controlled aerial robotic systems would be able to implement that hierarchy that OSHA hierarchy of fall protection by eliminating the hazards of falling, or the risk of falling, by having the operator stay on the ground, having the aerial robotic system go up and do the work in their place.
Again, some of the benefits of this, in addition, are that you’re able to replace human judgment for science. You could increase the safety and the productivity, possibly reduce the cost and the labor, and eventually provide measured standardized quality with sustainable benefits.
One example, and you’re seeing here on this slide, is a picture of an aerial robotic system that’s spray painting the side of a building. I painted houses working my way through college, and what you learn over time is the proper distance to stay from the wall with your spray head when you’re painting to get the optimal amount of paint on the wall. That’s actually called transfer efficiency ratio.
What you learn is, if you move from the sunny side of the building to the shady side of the building, you move that distance because the variable has changed. You learn it’s a more humid day, you move that distance. These were all things that were intuitive to a painter that they learn over time. An average painter can cover 400 square feet per hour, most novice painters start out painting 200 square feet per hour, but expert painters paint 600 square feet per hour.
A robotic system would be able to do this automatically. The sensors on, they would be able to say “here’s the barometric pressure, here’s the relative humidity, here’s the temperature, here’s the surface temperature of the structure being painted, here’s the viscosity of the paint, here’s the characteristics of the coating material.” And then adjust for it. It’s a simple look up table that software can utilize very easily.
This brings to mind the legend of John Henry of Man versus Machine. It’s interesting that this picture on this slide is from Bing for Labor Day. They have an image of a lot of people rappelling down a structure painting it. But the legend of John Henry is that in the late 1920s, I’m sorry, in the 1870s, at the Chesapeake and Ohio Railway Bend there was a tunnel that needed to be built for the railroad to go through the mountain around the bend of the river.
John Henry was the best tunnel digger at the time, and it was a man versus machine, and the legend is that John Henry cut through the mountain in the same amount of time that a machine did, but at the end of it he grabbed his heart and fell over dead. We all know that it’s very easy now for a machine to do much more physical labor that a person can do. This trend is not going to change, it’s just going to continue to get larger, and larger.
Again, robotic systems are awesome, they can do a lot of really cool, a lot of really good things, let’s use them for making the world safer.
Questions & answers
Q: Robert, I understand companies like Uber and Airbnb are software companies that don’t really own the cars or apartments they make money from, is there a similar model in the robotic space or for robotic software companies?
A: Yes, somewhat. You’ll see it with a lot of the apps in the App Store for example. There’s somebody that’s writing an app for a phone, but they don’t own the phone, and that’s one of the incredibly beautiful things about software is you write it once, and it can be installed millions of places. It doesn’t cost anymore, it’s bits and bytes, it’s not like manufacturing a chair, if you sell that chair 10 times, you have to physically put together 10 chairs.
This is, you sell 100,000 copies of software, they’re copies, there’s no additional, incremental cost for them. The software field has been using this for a long, long time, and was co-opted by Airbnb and Uber, but it’s similar. You’re utilizing, or piggybacking off of existing hardware by using the value of software. How this relates to robotics is it with robotics there does have to be a physical robot, but the software can do more and become better.
Again, I take it back to the example of the phone, and/or your desktop PC, that you’re continually getting updates on your software to make it better, to add functionality, to add features, while the hardware stays the same. You can still have that one piece of hardware doing one task, but maybe now with software upgrades that one piece of hardware can do two, or three, or four tasks.
Q: Robert, how do you see robotic software changing in the next three to five years?
A: It goes back to the presentation with Moore’s Law of hardware processing speed doubling every 18 months. As the value is unlocked, people understand the huge benefits, especially companies understand the benefits of software, it’s just going to become more and more ubiquitous, and more and more resources are going to be poured into it. As we talk about with industry 4.0, a lot of companies are looking to move from analog to digital, they’re looking to do that with information and data. That’s part of the pull, or the push behind big data, because you can unlock a huge amount of value that you didn’t have before you had that data.
As robotic started entering into tasks within the workforce, and do things, and as we start measuring things with various sensors, all that requires additional software.
Q: Are you concerned at all that these new robotic systems will take jobs from workers?
A: Not really. Actually, people are very concerned about that, and a lot of people have talked to this issue, but looking historically, at one point in time, a large percentage of the population work on farms, and now there are tractors that do a lot of the work, and we’re able to sustain our population, that’s with a much, much, much smaller number of workers. The same thing was said with computers when they first came out, typists, people on typewriters are going to lose their jobs, and they did, certainly, yes, most of them, but look at the productivity that that enabled.
There’s going to be hard times, there’s going to be a lot of transition, but the robotic systems, I believe, are going to greatly enhance the quality of life of people, and turn this from take a lot of the jobs that were menial, back-breaking jobs and turn them into jobs that are more cerebral, or utilize skills or human cognitive processing.
Not to mention the just enormous backlog of things that need to be done. For example, in infrastructure. Here in the United States, there are so many thousands of bridges that are below standard, and infrastructure that has been delayed, just even the power grid, the number of power transmission towers, most of those were built back in the 1950s and 1960s when the power grid was constructed. They need to be maintained and/or replaced. As we create robotic systems that construct things, and move things, and do things, and paint things, or do stuff like that, do things like that, that’s going to be able to take that backlog out of that, into actual work, that’s going to require additional people to do.
I can actually see this being something that would increase the jobs for people, because more work is going to need to be done.
Q: Can you talk a little bit on the security side of robotic software? How are companies ensuring that robots can’t be hacked by bad guys?
A: That’s absolutely a fantastic question. Given that robotic systems now are becoming more and more networked, and more and more interconnected, it’s more and more relevant. While in the past, robotic systems for the assembly line, for example, might have been what’s called offline, meaning they were self-contained and didn’t transmit any data outside of their own functionality, that’s changing rapidly.
Security is becoming even more and more critical. That’s something that companies need to pay close attention to, and we’re reminded with that every day with all the hacks that we see. This comes back to software once again, because how do you prevent the hacks with software? How are the hacks accomplished with software?
You have to make sure that you have well-written software that prevents this from happening in the first place, whatever possible.