Research

Background

The Sixth Science, Technology, and Innovation Basic Plan has identified the realization of "Society 5.0" as an essential direction for developing science and technology, in which cyber and physical technologies are integrated to achieve both economic development and solutions to social issues. In such a society, network systems in which objects are cyber-connected to each other through communication and sensing are expected to grow in scale. In addition to smartphones, PCs, and tablets, various devices such as automobiles, home appliances, houses, drones, and robots will be added as objects, which are expected to make the social infrastructure more sophisticated and efficient. The management of such large-scale systems requires "scalability," in which the load does not depend on the size of the system, and "robustness," in which the system is resistant to partial failures. A framework for "autonomous decentralized control" of objects using only local information obtained through communication and sensing is indispensable to realize such a system.

Outline

To develop fundamental management technology for large-scale systems, we have researched the design of autonomous decentralized control systems from theoretical and applied perspectives. In theoretical research, we have developed a theoretical system that captures the essence of the problem using mathematical tools such as optimization, graph theory, swarm theory, and manifold theory, with control theory as the core. In applied research, we have expanded the practical potential of the developed theory through the verification of swarm robots and applications to electric power and transportation systems in collaboration with companies and overseas researchers. We have promoted interdisciplinary activities spanning control engineering to robotics, electricity, transportation, and information engineering. In the future, we will create technological innovations to solve various social issues toward realizing the SDGs by integrating technologies from other fields (neural networks, optimal transport theory, quantum computation, quantum communication, etc.) with autonomous decentralized control technology as a core technology. Furthermore, we aim to develop a platform for using such technologies to realize a new society where diverse agents, including humans, machines, and the environment, cooperate to create value.

Plans

Control System Design for Groups of Agents with Heterogeneous Structures: Toward Cooperation among Machines, People, and the Environment

Most previous control theory research on large-scale systems has assumed that the agents have identical properties. This assumption allowed the design of scalable autonomous decentralized control systems because the entire system could be reduced to a "representative model of agents plus a network model. However, this assumption does not hold for systems in which agents with entirely different characteristics and structures, such as machines (robots, drones, cars), people, and environments (external sensors, buildings, roads), are mixed and scalable design by model reduction is not possible. This research aims to develop a scalable control system design method for heterogeneous agents and to realize a platform for cooperation among agents with entirely different structures, such as machines, humans, and environments. This will provide a technological foundation that can be applied to engineering and social sciences, such as developing robust infrastructures with diverse systems (SDG 9) and systematic organizational coordination in a diverse society (SDG 8).

Establishment of distributed control technology for mobility and energy systems

In the smart cities that will be realized by Society 5.0, it will be necessary to combine various forms of mobility and energy to achieve both personal mobility and social sustainability. To achieve this, one-way car-sharing services for electric vehicles and MaaS are being promoted to diversify the means of transportation. Since these systems are composites of elemental systems of entirely different quality, model reduction is impossible, making scalable design difficult. This research aims to establish a scalable design method for distributed control of mobility and energy systems and to realize a platform that induces actions based on the principles of Society 5.0. This will contribute to promoting clean energy deployment (SDG 7) and providing sustainable transportation systems (SDGs 9 and 11).

Real-time positioning and control by mobile swarms in non-GPS environments

Real-time positioning (acquisition of position coordinates) is routinely used in smartphones and car navigation systems and is an essential elemental technology for automating moving vehicles. In particular, highly accurate positioning is necessary to automatically control multiple drones and robots (groups of moving objects) for surveying and transportation. Positioning is mainly performed outdoors using GPS or in laboratories using external devices such as motion capture. It isn't easy to obtain highly accurate positioning in non-GPS environments such as indoors, in tunnels, and under bridge piers, where GPS is unavailable. This research aims to develop a technology to achieve highly accurate real-time positioning in non-GPS environments through distributed and cooperative positioning and control of a swarm of moving objects. This will enable the construction of a "positioning platform," a framework for obtaining location information regardless of location, and thereby contribute to automated driving of automobiles, construction machinery, and agricultural machinery (SGDs 8) and infrastructure inspections by drones (SGDs 9).

Results

Design theory of optimal distributed controllers for large-scale networked systems

In existing research, distributed controllers were designed only for specific tasks (control objectives), and the problem is that controller design requires trial-and-error and that other controllers with high performance may exist. Therefore, our team has developed a general-purpose method to design optimal distributed controllers systematically, regardless of the task. Precisely, first, the degree of accomplishment of various tasks is calculated as the deviation from the target set D, V = dist((x1, . . . , xn), D) (where (x1, . . . , xn) is the state sequence of the element system, and dist is the distance function). Next, we theoretically proved that an optimal decentralized controller can be designed by taking the sum of VC = dist(xC, projC (D)) (where proj is a projection operator) projected onto the clique C of the graph (a complete subgraph). He was the first in the world to show that the control engineering concept of "distributed control" is essentially characterized by the graph theory concept of "cliques.

Major results

K.Sakurama.Clique-based distributed PI control for multiagent coordination with heterogeneous, uncertain, time-varying orientations. IEEE Transactions on Control of Network Systems, Vol. 7, No. 4, pp. 1712– 1722, December 2020
K. Sakurama and H. Ahn. Multi-agent coordination over local indexes via clique-based distributed assignment. Automatica, Vol. 112, p. 108670, February 2020
K. Sakurama, S. Azuma, and T. Sugie. Multi-agent coordination to high-dimensional target subspaces. IEEE Transactions on Control of Network Systems, Vol. 5, No. 1, pp. 345–358, March 2018
K. Sakurama, S. Azuma, and T. Sugie. Design theory of distributed controllers via gradient-flow approach. In Emerging Applications of Control and System Theory, chapter 23, pp. 313–325. Springer, 2018 (Book)
K. Sakurama, S. Azuma, and T. Sugie. Distributed controllers for multi-agent coordination via gradient- flow approach. IEEE Transactions on Automatic Control, Vol. 60, No. 6, pp. 1471–1485, June 2015

Design Theory of Distributed Controllers for Autonomous Mobile Agent Swarms

Autonomous mobile agents" are cyber-physical systems with a wide range of applications, in which autonomous agents are connected through communication and sensing to perform tasks. Examples include mapping and warehouse management by a swarm of robots, building inspection and package transport by a swarm of drones, guidance of a group of people, and formation driving of self-driving vehicles. Since agents observe each other's relative positions through sensing, the information obtained depends on their own position and posture, as well as the type and performance of the sensors. In existing research, the controller had to be redesigned every time a new sensor was considered because it was targeted at a specific sensor. Therefore, our team has established a systematic design method for distributed controllers based on relative observations independent of sensor type. Specifically, first, to treat relative observation information uniformly regardless of sensors, we expressed x[i] ∈ {M x: M ∈ M} by a transformation set M . Next, we derived the necessary and sufficient condition for the feasibility of the task of the target set D to be the conditional expression D = orb(M) (where orb is an orbit), and designed a distributed controller based on optimal relative observations. Furthermore, we demonstrated the possibility of various applications, including a method for constructing a common coordinate system that does not depend on communication, and demonstrated its effectiveness in experiments using a mobile robot. These results show for the first time in the world that the relationship between "relative observation" and achievable "tasks" in control can be characterized by the "orbit" concept of group theory.

<Experiment of target assignment by ID-free control＞

Major results

K. Sakurama. Unified formulation of multi-agent coordination with relative measurements. IEEE Transactions on Automatic Control, Vol. 66, No. 9, pp. 4101–4116, September 2021
K. Sakurama and T. Sugie. Generalized coordination of multi-robot systems. Foundations and Trends in Systems and Control, Vol. 9, No. 1, pp. 1–170, 2021 (Book)
K. Sakurama, S. Azuma, and T. Sugie. Multi-agent coordination via distributed pattern matching. IEEE Transactions on Automatic Control, Vol. 64, No. 8, pp. 3210–3225, August 2019

Application to Smart Grid Distributed Control

In the "smart grid," a next-generation power system equipped with smart meters, demand-side power control, or "demand response," is expected to solve power problems such as power shortages and unstable power generation from renewable energy sources. Each utility receives information on its customers' electricity consumption through smart meters, determines the amount of price and incentive adjustments, and transmits this information. Our team proposed a decentralized control method that achieves demand response by exchanging information among smart meters without burdensome centralized control. Technically, it is a distributed solution of a bounded optimization problem through communication, and the necessary and sufficient condition for the imbalance to be eliminated is that the network is a strongly connected graph. Furthermore, we analyzed the stability of the network under realistic conditions where iterative calculations are terminated and proposed a method for masking privacy data, such as electricity consumption. The results are scalable to the number of houses and are expected to be a practical technology for solving future energy problems.

Major results

K. Sakurama. Control of large-scale cyber-physical systems with agents having various dynamics. IEEE Transactions on Big Data, Vol. 6, No. 4, pp. 691–701, December 2020
K. Sakurama and H. Ahn. Network-based distributed direct load control guaranteeing fair welfare maximization. IET Control Theory & Applications, Vol. 13, No. 17, pp. 2959–2968, November 2019
K. Sakurama. Distributed flow network control with demand response via price adjustment. Neurocomputing, Vol. 270, pp. 34–42, December 2017
K. Wada and K. Sakurama. Privacy masking for distributed optimization and its application to demand response in power grids. IEEE Transactions on Industrial Electronics, Vol. 64, No. 6, pp. 5118–5128, June 2017
K. Sakurama and M. Miura. Communication-based decentralized demand response for smart microgrids. IEEE Transactions on Industrial Electronics, Vol. 64, No. 6, pp. 5192–5202, June 2017
K. Sakurama and M. Miura. Distributed constraint optimization on networked multi-agent systems. Applied Mathematics and Computation, Vol. 292, pp. 272–281, January 2017
M. Miura, Y. Tokunaga, and K. Sakurama. Graphical and scalable multi-agent simulator for real-time pricing in electric power grid. Artificial Life and Robotics, Vol. 21, No. 2, pp. 181–187, June 2016

Application of Smart Mobility to Distributed Control

Smart mobility," a next-generation transportation system that utilizes communication and sensing by smart phones and connected cars, is expected to solve transportation problems such as traffic congestion and accidents, and create a new form of service (MaaS: Mobility as a Service). We have conducted research on smart mobility using mathematical and information sciences to promote multifaceted studies. In particular, to solve the problem of uneven distribution of vehicles in one-way (drop-off) car sharing services, we proposed a method to determine how to reallocate vehicles by staff and how to guide customers with dynamic fees using various mathematical tools (consensus control, sparse control, Wasserstein distance, event-driven distributed optimization, DC planning, etc.). The proposed method determines how to guide customers by using various mathematical tools (consensus control, sparse control, Wasserstein distance, event-driven distributed optimization, and DC planning). Among them, we derived a decentralized pricing rule that determines the optimal dynamic fare based on local communication among stations, and verified its effectiveness using data from the Ha:mo TOYOTA car sharing service in Toyota City. Next, we conducted an international joint study on decentralized control of traffic systems to alleviate traffic congestion. We proposed a distributed model predictive control method in which traffic signals communicate with neighboring vehicles and other traffic signals to determine the green light time. This result shows the role of mathematical and information sciences in the mobility society, and directs future research.

Major results

K. Sakurama, K. Kashima, T. Ikeda, N. Hayashi, K. Hoshino, M. Ogura, and C. Zhao. System-control-based optimization of one-way car-sharing services. In Advanced Mathematical Science for Mobility Society. Springer. (In Press) (Book)
K. Sakurama. Optimal control and station relocation of vehicle-sharing systems with distributed dynamic pricing. IEEE Open Journal of Intelligent Transportation Systems, Vol. 4, pp. 393–405, May 2023
N. Hayashi and K. Sakurama. Communication-aware distributed rebalancing for cooperative car-sharing service. IET Control Theory & Applications, Vol. 17, No. 7, pp. 850–867, April 2023
C. Zhao, K. Sakurama, and M. Ogura. Optimization of buffer networks via DC programming. IEEE Transactions on Circuits and Systems II: Express Briefs, Vol. 70, No. 2, pp. 606–610, February 2023
V. H. Pham, K. Sakurama, S. Mou, and H. Ahn. Distributed control for an urban traffic network. IEEE Transactions on Intelligent Transportation Systems, Vol. 23, No. 12, pp. 22937–22953, December 2022
T. Ikeda, K. Sakurama, and K. Kashima. Multiple sparsity constrained control node scheduling with application to rebalancing of mobility networks. IEEE Transactions on Automatic Control, Vol. 67, No. 8, pp. 4314–4321, September 2022

Page updated

Google Sites

Report abuse