Final Report Summary - DEMAND (DECENTRALIZED MONITORING AND ADAPTIVE CONTROL FOR NETWORKED DYNAMICAL SYSTEMS)
Networked Cyber-Physical Systems typically consist of large number of networked/interconnected physical and computational elements/subsystems. Decision making in such large-scale, complex and uncertain systems is a challenging problem. The main objective of the project was to develop new methods, algorithms and techniques for the tasks of decentralized estimation, detection and adaptive control for networked dynamical systems based on novel game-theoretic and consensus methodologies, with a special focus on applications to mobile wireless sensor networks.
The main research results have been achieved in developing novel methodologies, algorithms and tools for: decentralized overlapping tracking control based on decomposition of large-scale systems, decentralized overlapping system identification, decentralized adaptive target tracking, distributed stochastic approximation, distributed macro-calibration and time synchronization in sensor networks, robust fault detection in networked systems, distributed reinforcement and supervised learning, stochastic extremum seeking control on manifolds with applications to adaptive distance-based synchronization of oscillators and rigid bodies, obstacle avoidance control, including several practical applications to wireless sensor networks.
The achieved scientific results completely fulfill the goals of the project. All the planned milestones have been fully attained. Some of the obtained results went even beyond the defined project objectives. Overall, it can be concluded that the project research results are excellent, resulting in outstanding number of journal and conference publications (14 papers in top international journals and 13 peer reviewed international conference papers).
The conducted research is expected to have high technological and scientific impact on numerous real-world applications in general areas of Sensor Networks (SN), the Internet of Things (IoT) and Networked Cyber Physical Systems (NCPS), which have attracted recently a great deal of attention, due to advances in high speed networks, together with the low cost and powerful computational capabilities of sensing devices. More specifically, methods and algorithms for decentralized system decomposition, estimation, identification and control are of crucial importance for applications in Smart Grids, Smart Societies, Industrial Internet of Things IoT (IIoT), Social Networks, Biological Systems, surveillance using formations of Unmanned Aerial Vehicles (UAVs), etc. The extremum seeking control has properties which render it attractive for many application areas such as brake system control, real-time optimization of bioprocesses and flow control. Macro-calibration and time synchronization in networked sensing systems represent crucially important challenges for the wide deployment of the mentioned technologies. Another important problem in multi-device NCPS and IoT systems is that, as the number of devices increases, so does the possibility of faults which, when undetected, can lead to several flavors of detriments: from mild inconveniences in, e.g. HVAC systems (poor air quality) to disruptive ripple effects in, e.g. power systems (extended blackouts). Hence, the importance of fault detection and diagnosis in these systems is undisputable. Furthermore, the agents in NCPS and IoT typically operate in dynamically changing and complex environments so that they need to learn new behaviors online, using reinforcement learning methods. Finally, methods for supervised learning represent crucially important ingredient in Data Science, the recently exploded scientific field dealing with extracting knowledge or insights from data in various forms. In general, the conducted research belongs to several most challenging areas in the actual domain of convergence of Computers, Communications and Control. The extreme actuality and importance of the proposed and conducted research is beyond any doubt.
The fellow was engaged in many activities related to the transfer of knowledge to the Host Institution (Innovation Center, School of Electrical Engineering, University of Belgrade – ICEF) and to the dissemination of the project results. He established the IoT Lab within ICEF with the purpose of hosting many other research, innovation, technical and teaching/training projects in the field, and strengthening existing and initiating new collaborations with academia and industry at both national and international levels. He has been lecturer of several graduate courses at the University of Belgrade, a co-lecturer of two international graduate courses, a co-organizer of an international workshop, and he gave visiting lectures and seminars at several other national and international universities. During the project, the fellow has attended 11 international and national conferences. He was a keynote speaker at one international symposium in Belgrade. He is also participating in another national research project, and in one industrial innovation project. The fellow is finalizing several research and innovation project proposals on the currently cutting-edge topics, with several national and international academic and industrial participants. He was elected as a member of the IEEE Control Systems Society Conference Editorial Board which further strengthens his international connections. He has also been engaged in other outreach activities aimed at wider public such as: keynote talks on round tables at national conferences, presentations at the Belgrade Fair of Technique and Technical Achievements, and interviews for the press.
The fellow is currently full-time employed by ICEF as a Research Assistant Professor. He plans to stay at that position at least until the end of the academic year since he is currently engaged in several teaching activities and projects. Until that point he will decide whether to stay there or to accept some other position offer.
The main research results have been achieved in developing novel methodologies, algorithms and tools for: decentralized overlapping tracking control based on decomposition of large-scale systems, decentralized overlapping system identification, decentralized adaptive target tracking, distributed stochastic approximation, distributed macro-calibration and time synchronization in sensor networks, robust fault detection in networked systems, distributed reinforcement and supervised learning, stochastic extremum seeking control on manifolds with applications to adaptive distance-based synchronization of oscillators and rigid bodies, obstacle avoidance control, including several practical applications to wireless sensor networks.
The achieved scientific results completely fulfill the goals of the project. All the planned milestones have been fully attained. Some of the obtained results went even beyond the defined project objectives. Overall, it can be concluded that the project research results are excellent, resulting in outstanding number of journal and conference publications (14 papers in top international journals and 13 peer reviewed international conference papers).
The conducted research is expected to have high technological and scientific impact on numerous real-world applications in general areas of Sensor Networks (SN), the Internet of Things (IoT) and Networked Cyber Physical Systems (NCPS), which have attracted recently a great deal of attention, due to advances in high speed networks, together with the low cost and powerful computational capabilities of sensing devices. More specifically, methods and algorithms for decentralized system decomposition, estimation, identification and control are of crucial importance for applications in Smart Grids, Smart Societies, Industrial Internet of Things IoT (IIoT), Social Networks, Biological Systems, surveillance using formations of Unmanned Aerial Vehicles (UAVs), etc. The extremum seeking control has properties which render it attractive for many application areas such as brake system control, real-time optimization of bioprocesses and flow control. Macro-calibration and time synchronization in networked sensing systems represent crucially important challenges for the wide deployment of the mentioned technologies. Another important problem in multi-device NCPS and IoT systems is that, as the number of devices increases, so does the possibility of faults which, when undetected, can lead to several flavors of detriments: from mild inconveniences in, e.g. HVAC systems (poor air quality) to disruptive ripple effects in, e.g. power systems (extended blackouts). Hence, the importance of fault detection and diagnosis in these systems is undisputable. Furthermore, the agents in NCPS and IoT typically operate in dynamically changing and complex environments so that they need to learn new behaviors online, using reinforcement learning methods. Finally, methods for supervised learning represent crucially important ingredient in Data Science, the recently exploded scientific field dealing with extracting knowledge or insights from data in various forms. In general, the conducted research belongs to several most challenging areas in the actual domain of convergence of Computers, Communications and Control. The extreme actuality and importance of the proposed and conducted research is beyond any doubt.
The fellow was engaged in many activities related to the transfer of knowledge to the Host Institution (Innovation Center, School of Electrical Engineering, University of Belgrade – ICEF) and to the dissemination of the project results. He established the IoT Lab within ICEF with the purpose of hosting many other research, innovation, technical and teaching/training projects in the field, and strengthening existing and initiating new collaborations with academia and industry at both national and international levels. He has been lecturer of several graduate courses at the University of Belgrade, a co-lecturer of two international graduate courses, a co-organizer of an international workshop, and he gave visiting lectures and seminars at several other national and international universities. During the project, the fellow has attended 11 international and national conferences. He was a keynote speaker at one international symposium in Belgrade. He is also participating in another national research project, and in one industrial innovation project. The fellow is finalizing several research and innovation project proposals on the currently cutting-edge topics, with several national and international academic and industrial participants. He was elected as a member of the IEEE Control Systems Society Conference Editorial Board which further strengthens his international connections. He has also been engaged in other outreach activities aimed at wider public such as: keynote talks on round tables at national conferences, presentations at the Belgrade Fair of Technique and Technical Achievements, and interviews for the press.
The fellow is currently full-time employed by ICEF as a Research Assistant Professor. He plans to stay at that position at least until the end of the academic year since he is currently engaged in several teaching activities and projects. Until that point he will decide whether to stay there or to accept some other position offer.