As of now, "cyberphysic systems" is something that even the top leaders in different ministries have a vague understanding of. Even the term in Russian is just a calque from English. Many inquire on what is the difference between blockchain and the Internet of Things. There is an official definition: a network of physical and computational objects that are interconnected and designed as a single system, that is organized in the framework of a common basic cyberphysic model and can adapt to changes in the real world. Such systems can be controlled and adaptive; there are several key problems to them: cyber security, scalability, reliability. These systems are applied in most fundamental areas of human activity — communications, consumer electronics, power industry, infrastructure, healthcare, medicine, new manufacturing technologies — like robotics and intelligent transport systems.
If one looks at the cyberphysic's developmental curve, we face this change when we’ve overgrown the Industrial Age and entered the era of informational technologies. Now we see a new change, back to the physical world, when we've developed all the necessary informational technologies and it is time to apply them. So, is there anything of essence to this new term? When people talk about cyberphysic systems, they imply an incorporation of different technologies — computational, information, control and those of the physical world. It is all about co-design in software development, as well as convergence — different technological systems evolving to solve similar tasks. The physical reality now has its digital copy: when a new product is being developed, everything, even the idea of its design, starts with a digital model.
Trends
I believe that software development (programs for programs) and developing programs for hardware are separate issues, as the requirements are different. Another trend is transferring computations out of servers. If cloud computations were a popular topic once, now it's the so-called fog computing where you don't use a big powerful server, but everything is computed right on site using integrated controllers that can work under harsh conditions and retain their efficiency even when computational capacities are restricted.
There are also action networks, where information itself partakes in decision making when controlling physical objects. Here, the requirements to security and reliability are totally different. Internet banking is my favorite example. Everyone uses it, sometimes failures occur, when a transaction goes the wrong way. It's regrettable, but it can be fixed. Still, if there's a failure or a hack of an auto-piloted car that then drives into a crowd of people, you can't possibly fix it afterwards, and the consequences would be fatal.
A trend that has to do with smartphones is SDX — when the worth of something is mostly defined by software, not hardware. In smartphones, what's important to you is not cameras or g-meters, but the apps you download. Another example is the modern concepts in car manufacturing, Tesla, for instance. You had your plain car and then it becomes partially autonomous — just because you've paid for new software. You didn't have to replace the sensors or go to the service for that.
Challenges
Naturally, the capacity for these services has to be introduced on the stage of design. Several examples of the "smart world" technologies like cars communicating with infrastructure and other cars scale from a single car to controlling traffic flows of a whole city or a highway. Speed limitations, traffic lights, priority grades — all of that can be balanced using virtual transport technologies. For instance, last year in Europe there was an event dedicated to platooning systems, when cars followed each other in a line. So what's the catch? The cars went in a line, the air flow's resistance was considerably lower, there were fewer changes in speed, the fuel conservation went up by 5 to 20%, CO2 exhausts were 10% less. The system's reaction to emergency braking was 25% better than that of a live person's would. Also, the traffic flows and such were optimized. The startup for developing this technology gathered 30 million dollars last year.
Another great example is a "smart" manufacturing. Laptops and tablets are really versatile instruments for mental activities — now robots are to become the same for manual labor. There is the KUKA Company that introduces its robots to plants. The Daimler Company uses them in manufacturing automated gearboxes. A person can accidentally put his hand into the gearbox that's being manufactured. The robot can notice it and avoid harmful contacts.
Per aspera ad astra
The process of introducing a technology — from design to commercial success — is really challenging, and there can be lots of problems. A new technology starts on the stage of ideas and fundamental research, then goes applied research, then product design and certification. And there are different players on each stage — companies that deal with technology transfer, startups and the government. Still at some point the organizations that developed the technology don't know how to develop it further, and business entities still don't get how to commercialize it. This is a turning point — something like a Valley of Death that few projects pass. Some have to go through it again after their initial failure. In essence, that is what we call innovations.
Sometimes, the problems are totally unexpected — like with Google Glass. The problem was that wearing glasses proved to be uncomfortable, and no PR campaigns could help it. Also, Google Glass wasn't an essentially new experience, so many were disappointed with it. Another example is Cherepanov's steam train. The invention "died", because the manufacturers didn't believe in it. No one tried developing it further so it would consume less firewood; the project was just closed. It happens really often, and that's a real problem for the development of technologies.
