What are cyberphysical systems, why are they so relevant today, and what role does artificial intelligence play in their development?
The study of the various properties of information technology systems in terms of the interaction of their physical and digital components is a new and relevant area of modern science of cyberphysical systems [1].
The main components of any cyberphysical system are (Fig. 1):
- ( );
- ( , , , .);
- ( , .);
- ( XR ).
.1 , .
, , . , , .. .
, , , : .
(Cyber-physical systems, CPS) — . : CyPhyLab , MIT media lab, , Bosch, Toshiba PTC.
CPS - , , , .. “”. , CPS — , , . , , , CPS (.2).
.2 .
, , 4.0.
Acatech, 4.0, , : , . , , “ ”, . — , , 2011 . .
.3 , .
, CPS, , . — .
- 3- , , , , , . , , . , .
. 4
, .
. 5 .
. [3]. , [4–6]. , , , . , , . , , .
. habr.
. 6 . ( , HighLoad++ 2018).
, , 2- . . — , .
. 7 ( , HighLoad++).
. 8 .
(), . SpaceX Starlink. , .
— . . , , . , - , .
. 9 VR .
TZUP, PHYGITALISM , . ( ), .. . . , , . . .
. 10 .
CPS
Fig. 11 Visualization of the simulation of technological processes in the digital twin of the plant.Another important issue is ensuring the security of CPS - the creation of cyber-physical attack protection systems has become a new topical area in the field of information security. You can familiarize yourself with the thoughts on this subject in the next note .Let's look at a practical case called PlantSimimplemented by the PHYGITALISM team. The main objective of the project was to create an anomaly detection system in the communication channels between the PLC (programmable logic controller), which controls the technological processes at the refinery, and the engineering monitoring system (SCADA). To search for anomalies in the control channels, it was proposed to use a recurrent neural network [7].
Fig. 12 Block diagram of the cyberphysical system under consideration.As data for training such a network, synthetic data obtained from the digital double of the plant is used. The advantage of this approach, compared with the traditional solution based on the logical rules of control of communication channels, lies in the variety of situations that the artificial intelligence system can detect. With the help of a digital double of the plant, it is possible to consider even such rare but potentially dangerous situations as a reactor explosion. And of course, an important factor is the speed with which new data can be obtained for training the system.AI strengthens the sensory organs of CPS
Fig. 13 Google project soli sensor.For the best quality of functioning of the CPS, it is necessary to have as much information as possible about the state of the environment in which such a system operates, and to determine as accurately as possible the processes occurring in its components. All kinds of sensors and sensors serve as sensory organs for CPS. The better the system “sees” what is happening with it and around it at a given moment in time, the better will be the quality of its decisions on managing available resources.For example, in the context of the previous section, the more sensitive thermostat is installed in the tank for the transport of explosive liquids (provided that the thermostat is insensitive to noise emissions), the sooner in case of a heating failure the security system can detect the danger and prevent it.
Fig. 14 Using project soli to control devices using gestures.As a more vital example, you can consider the recognition of subtle human gestures using a wi-fi signal to control gadgets.And here, artificial intelligence can help with improving the performance of existing sensors or with the creation of new sensors. This is clearly demonstrated by the Google Project Soli sensor.The basic idea here is not new: wi-fi routers are transmitter-receivers of electromagnetic waves. Passing through space, an electromagnetic wave is reflected from various objects and returns back to the device. If an object begins to move, then its speed is added or subtracted with the speed of the electromagnetic wave, which causes a shift in the frequency and length of the reflected wave. This phenomenon is called the Doppler effect , and we can use it for detecting gestures, and not just using a wi-fi router . However, not all so simple.Firstly, ordinary routers operate at frequencies of the order of 5GHz. Such a frequency may be insufficient for a good resolution of the classification algorithm, i.e. close in form, but still essentially different objects will be recognized as the same object. This problem is solved by increasing the operating frequency of the device (project soli has an operating frequency of 57 - 64 GHz).Secondly, in real tasks, it is required to recognize not only a specific gesture, but also to whom this gesture belongs, in one room there can be several people at once and all of them may have different priorities in the system. This and many other unannounced problems are solved by using modern deep learning algorithms [8]. This approach, for example, is used in the WiSee project .implemented by students and staff at the University of Washington.Since the work of neural network algorithms requires quite complex, but the same type of calculations, for their implementation in hardware there are special computing units, the so-called TPU (tensor process init) or otherwise - neuromodules, neuroprocessors. Such computing units can be optimized for a specific task, whether it is recognition of faces, voices or gestures. Just to solve the problem of recognition of gestures or spatial movements of small objects, a special TPU “Soli I-CB” was created, which can be built into various smart devices and gadgets.
Fig. 15 Neurochips project soli.Conclusion
Cyberphysical systems are an important component of the information age. The better we understand how to design, analyze, protect and improve such systems, the better, safer and more understandable we make the world around us. As you can see from the above examples, artificial intelligence can improve the performance of CPS at all levels, and the synergy of these two relevant areas (AI + CPS) will bring many interesting and unexpected improvements to our lives.If you are a undergraduate student or are interested in the topic of CPS, AI, XR technologies and the implementation of similar projects, then for this case we have a master's program “Augmented / Virtual Reality and Artificial Intelligence” based on the Faculty of Applied Informatics and Information Technology of the Moscow Aviation Institute, in collaboration with technology studioPHYGITALISM . The program is based on courses on artificial intelligence, cyberphysical systems, virtual and augmented reality, in the framework of which we will help you understand all the tricks of working with technology in practice.Waiting for you!Author
Vadim Kondrattsev,R&D Engineer PHYGITALISMLecturer at the Moscow Aviation Institutevadim@phygitalism.comitcentrmai@gmail.comSources1. Allgöwer, Frank & Sousa, João & Kapinski, James & Mosterman, Pieter & Oehlerking, Jens & Panciatici, Patrick & Panshai & Pansh & Ranhai , Akshay & Tabuada, Paulo & Wenzelburger, Philipp. (2019). Position paper on the challenges posed by modern applications to cyber-physical systems theory. Nonlinear Analysis. 34. 147–165. 10.1016 / j.nahs.2019.05.00.007.2. Paulo Tabuada, Cyber-Physical Systems: Position Paper3. Lee, Edward A. “The past, present and future of cyber-physical systems: a focus on models.”Sensors (Basel, Switzerland) vol. 15.3 4837–69. 26 Feb. 2015, doi: 10.3390 / s1503048374. RG Sanfelice. Analysis and Design of Cyber-Physical Systems. A Hybrid Control Systems Approach // Cyber-Physical Systems: From Theory to Practice / D. Rawat, J. Rodrigues, I. Stojmenovic. - CRC Press, 20165. Shao Z., Liu J. (2013) Spatio-temporal Hybrid Automata for Cyber-Physical Systems. In: Liu Z., Woodcock J., Zhu H. (eds) Theoretical Aspects of Computing - ICTAC 2013. ICTAC 2013. Lecture Notes in Computer Science, vol 8049. Springer, Berlin, Heidelberg6. EA Lee and SA Seshia, Introduction to Embedded Systems - A Cyber-Physical Systems Approach , http://LeeSeshia.org, 2011.7. Filonov, Pavel & Kitashov, Fedor & Lavrentyev, Andrey. (2017). RNN-based Early Cyber-Attack Detection for the Tennessee Eastman Process. ArXiv abs / 1709.022328. Qifan Pu, Sidhant Gupta, Shyam Gollakota, Shwetak Patel (2013) Whole-Home Gesture Recognition Using Wireless Signals // The 19th Annual International Conference on Mobile Computing and Networking (Mobicom'13)