On general control methodology for complex systems

<p>Control of complex systems is one of important files in complex systems, that not only relies on the essence of complex systems which is denoted by the core concept – emergence, but also embodies the elementary concept in control theory. Aiming at giving a clear and self-contained description of emergence, the paper introduces a formal way to completely describe the formation and dynamics of emergence in complex systems. Consequently, this paper indicates the Emergence-Oriented Control methodology that contains three kinds of basic control schemes: the direct control, the system re-structuring and the system calibration. As a universal ontology, the Emergence-Oriented Control provides a powerful tool for identifying and resolving control problems in specific systems.</p>

This paper discusses a kind of elementary methodology for the control of complex systems, called emergence-oriented control (EOC), and its key schemes: the direct scheme, the system restructuring scheme and the system calibration scheme. Firstly, the schemes are dependent on emergent behaviors and the emergence mechanisms producing those behaviors. The second is that an efficiency control relies on not only corresponding emergence mechanism, but also controlled components and interactions among them. As a universal ontology, the EOC provides a powerful tool for identifying and resolving control problems in specific systems.

Nowadays, modern approaches to complex system's control show the need for preparing strategies also for the atypical operational situations, like crisis, emergency, structural failures, starting and stopping of plants, etc. Situational control tries to handle the problem of the contradiction arising from the existence of the large number of situations and the limited number of used control strategies by grouping the possible situations into a treatable (finite) number of model classes of operational situations and assigning certain control algorithms to the defined control regimes. Anytime control algorithms fit into the frame of situational control concept. They aim to offer a tradeoff between resource consumption (including time and data as well) and output quality. In this paper, an anytime control methodology is presented which combines Takagi-Sugeno fuzzy modeling techniques with anytime processing. The proposed method can be used in situational control advantageously and is able to cope with atypical situations, time and resource insufficiency.

Complex systems are challenging to control because the system responds to the controller in a nonlinear fashion, often incorporating feedback mechanisms. Interdependence of systems poses additional difficulties, as cross-system connections enable malicious activity to spread between layers, increasing systemic risk. In this paper we explore the conditions for an optimal control of cascading failures in a system of interdependent networks. Specifically, we study the Bak–Tang–Wiesenfeld sandpile model incorporating a control mechanism, which affects the frequency of cascades occurring in individual layers. This modification allows us to explore sandpile-like dynamics near the critical state, with supercritical region corresponding to infrequent large cascades and subcritical zone being characterized by frequent small avalanches. Topological coupling between networks introduces dependence of control settings adopted in respective layers, causing the control strategy of a given layer to be influenced by choices made in other connected networks. We find that the optimal control strategy for a layer operating in a supercritical regime is to be coupled to a layer operating in a subcritical zone, since such condition corresponds to reduced probability of inflicted avalanches. However this condition describes a parasitic relation, in which only one layer benefits. Second optimal configuration is a mutualistic one, where both layers adopt the same control strategy. Our results provide valuable insights into dynamics of cascading failures and and its control in interdependent complex systems.

In systems development, it is necessary to think thoroughly about any emergent behaviors that might occur and create strategies to prevent the undesirable emergent behavior from occurring or at least mitigate them. Often dealing with that at later development stages is costly. Emergent behaviors are the main concern, particularly in the development of complex systems. However, identifying and predicting them is challenging in complex systems as it involves lots of components and interactions. Prior to making wise design decisions, developers need to elucidate and understand the emergent behaviors. For this reason, a conceptual model that highlights the presence of emergent behavior at an early stage is essential in the systems development. The conceptual model is useful in achieving higher coverage as it is correlated with the probability of identifying more emergent behaviors. Therefore, this paper goes through the existing work focusing on effort in modeling emergent behavior and presents perspectives on developing a conceptual model that is able to inform the presence of emergent behavior.

The semiotics of control and modeling relations in complex systems

We studied the ability of deep reinforcement learning and self-organizing approaches to adapt to dynamic complex systems, using the applied example of traffic signal control in a simulated urban environment. We highlight the general limitations of deep learning for control in complex systems, even when employing state-of-the-art meta-learning methods, and contrast it with self-organization-based methods. Accordingly, we argue that complex systems are a good and challenging study environment for developing and improving meta-learning approaches. At the same time, we point to the importance of baselines to which meta-learning methods can be compared and present a self-organizing analytic traffic signal control that outperforms state-of-the-art meta-learning in some scenarios. We also show that meta-learning methods outperform classical learning methods in our simulated environment (around 1.5-2× improvement, in most scenarios). Our conclusions are that, in order to develop effective meta-learning methods that are able to adapt to a variety of conditions, it is necessary to test them in demanding, complex settings (such as, for example, urban traffic control) and compare them against established methods.

Modeling and control of nonlinear systems using an Adaptive LAMDA approach

PEMFC involves a multi-physics coupling process which leads to some challenging difficulties for control strategy design and implementation. Among those, temperature regulation of PEMFC stack is essential to ensure the high operation efficiency, avoid thermal fatigue and prolong the stack lifespan [1, 2]. Several control strategies regarding to the temperature regulation have been carried out in recent years [3]. In [4], an extremum seeking control with a proportional-integral (PI) controller was proposed in order to achieve the optimal temperature regulation. In [5], the temperature control strategy was proposed for a PEMFC system with a water cooling system, where the voltage of a water circulation pump was manipulated through a PI controller and a water cooling fan was under the on/off modes. In [6], a more advanced fuzzy logic controller was applied to regulate the speed of the cooling fan and increase the convection ratio between PEMFC and atmosphere. However, the proposed control strategy is validated for a low-power PEMFC. Thus, the limitation of cooling capacity is not considered in those aforementioned control strategies [4,5,6]. When the generated heat from PEMFC is higher than the dissipated heat from the cooling system, the temperature of PEMFC will keep increasing. Recently, Chen et al. [7] considered the limited capacity of a water cooling system and current constraint in the PEMFC system and proposed three temperature control strategies, which developed a constrained model predictive control (MPC) strategy, and designed duty ratio split controller and current governor both based on the proportional-integral-derivative (PID) control in order to implement the trade-off between cooling capacity and power demand.

Nowadays, advanced controller design is called upon to guarantee the secure and reliable operation of power systems. To meet this requirement, this work proposed three control strategies: 1) robust coordinated transient stability and voltage regulation enhancement (CTSVRE) control of generator excitation systems, which enhances both transient stability and post-fault voltage regulation; 2) robust coordinated control of AC/DC power systems with overlapping decomposition, which coordinates both AC system control and DC system control from a systematic standpoint; 3) robust constrained control of generator excitation systems, which considers input saturation from the beginning of controller design. The designed controllers are tested by numerical simulation using PSCAD/EMTDC software. Their performances are compared with those of other controllers and evaluated using different cases. Results show that the proposed control strategies are quite effective.

The challenge we address in this work is the control of complex systems based on the multi-agent paradigm. Our focus is on identifying what specific questions arise when using the multi-agent paradigm in the control of complex systems context. We present here a solution under the form of an equation-free control architecture based on multi-agent model simulation. The results of the implementation on a free-riding problem in p2p networks demonstrate that the proposed architecture can control such a network. Our contribution is to identify key questions that rise when using the multi-agent paradigm in the context of control of complex systems, concerning the relationship between the model entities and the target system entities.

Coupling control and human factors in mathematical models of complex systems

This chapter focuses on the key difference between the two paradigms of crisis management described in Chapters 2 and 3. The traditional crisis planning process described in Chapter 2 depends on the ability of managers to predict and control multiple features of crisis situations: how organizations will behave during a crisis, how publics will respond to efforts to contain or repair the crisis, and how the company's actions will affect subsequent outcomes. The complexity view described in Chapter 3 substantially limits actors' ability to either predict or control all the interacting elements that make up a complex system.

Agent-based models (ABM) and their simulations have been used to study complex systems with interacting entities and to model multi-agent systems. Simulations are used to explore the dynamic consequences of these models. In many cases, the behaviours that are of interest are emergent ones that arise as a result of interactions between agents rather than the actions of any individual agent. In this paper, we propose a formalism for describing emergent behaviours at any level of abstraction based on the idea that event types can be defined that characterise sets of behavioural 'motifs'. This provides the basis for a method for studying the associations between multi-level behaviours in simulations. There are two categories of hypotheses that we seek to address with respect to an ABM and its simulations: - Hypotheses concerned with associations between emergent behaviours defined at various levels of abstraction. - Hypotheses concerned with the links between parameter sensitivity / initial conditions and emergent behaviours e.g. the ABM is sensitive to a parameter x because x predisposes the system or part of the system to exhibit a particular (emergent) behaviour.

The modern world is complex beyond human understanding and control. The science of complex systems aims to find new ways of thinking about the many interconnected networks of interaction that defy traditional approaches. Thus far, research into networks has largely been restricted to pairwise relationships represented by links between two nodes.This volume marks a major extension of networks to multidimensional hypernetworks for modeling multi-element relationships, such as companies making up the stock market, the neighborhoods forming a city, people making up committees, divisions making up companies, computers making up the internet, men and machines making up armies, or robots working as teams. This volume makes an important contribution to the science of complex systems by: extending network theory to include dynamic relationships between many elements; providing a mathematical theory able to integrate multilevel dynamics in a coherent way; providing a new methodological approach to analyze complex systems; and, illustrating the theory with practical examples in the design, management and control of complex systems taken from many areas of application.