· · · Institutional Repository

We consider the problem of designing controllers for spatially-varying interconnected systems distributed in one spatial dimension. The matrix structure of such systems can be exploited to allow fast analysis and design of centralized controllers with simple distributed implementations. Iterative algorithms are provided for stability analysis, H infin analysis and sub-optimal controller synthesis. For practical implementation of the algorithms, approximations can be used, and the computational efficiency and accuracy are demonstrated on an example.

A new prototype adaptive deformable mirror for future AO-systems is presented that consists of a thin continuous membrane on which push-pull actuators impose out-of-plane displacements. Each actuator has ±10μm stroke, nanometer resolution and only mW’s heat dissipation. The mirror’s modular design makes the mechanics, electronics and control system extendable towards large numbers of actuators. Models of the mirror are derived that are validated using influence and transfer function measurements. First results of a prototype with 427 actuators are also presented.

Multiagent systems are rapidly finding applications in a variety of domains, including robotics, distributed control, telecommunications, and economics. The complexity ofmany tasks arising in these domains makes them difficult to solve with preprogrammed agent behaviors. The agents must, instead, discover a solution on their own, using learning. A significant part of the research on multiagent learning concerns reinforcement learning techniques. This paper provides a comprehensive survey of multiagent reinforcement learning (MARL). A central issue in the field is the formal statement of the multiagent learning goal. Different viewpoints on this issue have led to the proposal of many different goals, among which two focal points can be distinguished: stability of the agents’ learning dynamics, and adaptation to the changing behavior of the other agents. The MARL algorithms described in the literature aim—either explicitly or implicitly—at one of these two goals or at a combination of both, in a fully cooperative, fully competitive, or more general setting. A representative selection of these algorithms is discussed in detail in this paper, together with the specific issues that arise in each category. Additionally, the benefits and challenges ofMARL are described along with some of the problem domains where the MARL techniques have been applied. Finally, an outlook for the field is provided.

Urban water supply networks are large-scale systems that transport potable water over vast geographical areas to millions of consumers. A safe and efficient operation of these networks is crucial, as without it living in today’s cities would be impossible. To achieve an adequate operation, these networks are equipped with actuators like pumps and valves, which are used to maintain water pressures and flows within safe margins. Currently, these actuators are controlled in a decentralized way using local controllers that only use local information and that do not take into account the presence of other controllers. As a result, water supply networks regularly experience pressure drops and interruptions of water supply when there is an unexpected increase in water demand. To improve performance the actions of the local controllers should be coordinated. Implementing a centralized control scheme is not tractable due to the large-scale nature of these networks. Therefore, this paper proposes the application of a distributed control scheme for control of urban water supply networks. The scheme is based on local model predictive control (MPC) strategies and a parallel coordination scheme that implements cooperation among the local MPC controllers. A simulation study based on a part of the urban water supply network of Bogotá, the capital of Colombia, illustrates the potential of the approach.

Irrigation canals are large-scale systems, covering vast geographical areas, and consisting of many interconnected canal reaches that interact with control structures such as pumps and gates. The control of such irrigation canals is usually done in a manual way, in which a human operator travels along the irrigation canal to adjust the settings of the gates and pumps in order to obtain a desired water level. In this paper we discuss how distributed model predictive control (MPC) can be applied to determine autonomously what the settings of these control structures should be. In particular, we propose the application of a distributed MPC scheme for control of the West-M irrigation canal in Arizona. We present a linearized model representing the dynamics of the canal, we propose a distributed MPC scheme that uses this model as a prediction model, and we illustrate the performance of the scheme in simulation studies on a nonlinear simulation model of the canal.

We consider multi-agent, or distributed, control of transportation networks, like traffic, water, and power networks. These networks typically have a large geographical span, modular structure, and a large number of components that require control. We discuss the necessity of a multi-agent control setting in which multiple agents control parts of the network. As potential control methodology we consider model predictive control (MPC) in a multi-agent setting. We first outline a framework for modeling transportation networks into subsystems using external variables and then discuss issues that arise when controlling these networks with multi-agent MPC. Several approaches to these issues are structured and discussed in terms of the outlined framework.

In this paper, we consider the problem of designing a distributed controller for a formation of spacecraft following a periodic orbit. Each satellite is controlled locally on the basis of information from only a subset of the others (the nearest ones). We describe the dynamics of each spacecraft by means of a linear time-periodic (LTP) approximation, and we cast the satellite formation into a state-space formulation that facilitates control synthesis. Our technique exploits a novel modal decomposition of the state-space model and uses linear matrix inequalities (LMIs) for suboptimal control design of distributed controllers with guaranteed performance for formations of any size. The application of the method is shown in two case studies. The first example is inspired by a mission in a low, sun-synchronous Earth orbit, namely the new Dutch-Chinese Formation for Atmospheric Science and Technology demonstration mission (FAST), which is now in the preliminary design phase. The second example deals with a formation of spacecraft in a halo orbit.

The refractive index of air varies a.o. with temperature, humidity, pressure and the CO2 concentration. Due to atmospheric turbulence this refractive index varies both in space and in time, leading to aberrations in images of light having passed though it. These aberrations limit the achievable resolution of optical telescopes such that the quality of their images is no longer diffraction limited. An adaptive optics (AO) system is a means to recover the diffraction limited quality of the images. This can be achieved a.o. by reflecting the incoming light on a deformable mirror (DM) that adapts its shape to the wavefront of this light such that some norm of the residual wavefront after reflection is minimal. In this thesis novel designs are considered for the DM and its control system. They are primarily aimed at the 8m class of telescopes in visible light, leading to a 200Hz controller bandwidth requirement and 6mm actuator spacing or 5000 actuators on a 500mm diameter DM.. To observe fainter celestial objects and/or increase the image resolution, optical telescopes are foreseen with primary mirrors of up to 40m in diameter. Therefore, the DM system design is aimed at extendibility to a larger number of Degrees Of Freedom (DOF), which is realized using a modular concept. Other drivers are low power consumption to prevent the need for active cooling systems and low production costs. The DM design is realized using electromagnetic reluctance type actuators that are connected to the DM's reflective membrane by a thin rod. Modules containing 61 hexagonally arranged actuators are manufactured using techniques suitable for mass-production. To generate the currents through the actuator coils, driver electronics are developed based on Pulse-Width Modulation (implemented in FPGAs) in combination with analog low-pass filters. Several prototype DMs are realized whose behavior is analyzed both statically and dynamically by comparing Wyko and laser vibrometer measurements with first principle models of the driver electronics, the actuators and the facesheet. To retain modularity of the system, a distributed control system architecture is foreseen in which each (group of) actuator(s) has its own controller that has a fixed computational power, communicates only to its neighbors and receives only a subset of the wavefront measurement data. The design of a distributed controller with good performance is complicated by the wavefront reconstruction step made necessary by the Shack-Hartmann wavefront sensor. Nevertheless, a distributed algorithm that combines wavefront reconstruction with adaptive prediction is shown in simulation to approximate the performance of a centralized finite impulse response (FIR) predictor/reconstructor and does not deteriorate as the number of DOFs increases.

Mobile robots that communicate and cooperate to achieve a common task have been the subject of an increasing research interest in recent years. These possibly heterogeneous groups of robots communicate locally via a communication network and therefore are usually referred to as robotic networks. Their potential applications are diverse and encompass monitoring, exploration, search and rescue, and disaster relief. From a research standpoint, in this thesis we consider specific aspects related to the foundations of robotic network algorithmic development: distributed estimation, control, and optimization. The word “distributed” refers to situations in which the cooperating robots have a limited, local knowledge of the environment and of the group, as opposed to a “centralized” scenario, where all the robots have access to the complete information. The typical challenge in distributed systems is to achieve similar results (in terms of performance of the estimation, control, or optimization task) with respect to a centralized system without extensive communication among the cooperating robots. In this thesis we develop effective distributed estimation, control, and optimization algorithms tailored to the distributed nature of robotic networks. These algorithms strive for limiting the local communication among the mobile robots, in order to be applicable in practical situations. In particular, we focus on issues related to nonlinearities of the dynamical model of the robots and their sensors, to the connectivity of the communication graph through which the robots interact, and to fast feasible solutions for the common (estimation or control) objective.

We consider the problem of designing distributed controllers for a class of systems which can be obtained from the interconnection of a number of identical subsystems. If the state space matrices of these systems satisfy a certain structural property, then it is possible to derive a procedure for designing a distributed controller which has the same interconnection pattern as the plant. This procedure is basically a multiobjective optimization under linear matrix inequality constraints, with system norms as performance indices. The explicit expressions for computing these controllers are given for both H infin or H 2 performance, and both for static state feedback and dynamic output feedback (in discrete time). At the end of the paper, two application examples illustrate the effectiveness of the approach.

Prior work on active aerodynamic load control (AALC) of wind turbine blades has demonstrated that appropriate use of this technology has the potential to yield significant reductions in blade loads, leading to a decrease in wind cost of energy. While the general concept of AALC is usually discussed in the context of multiple sensors and active control devices (such as flaps) distributed over the length of the blade, most work to date has been limited to consideration of a single control device per blade with very basic Proportional Derivative controllers, due to limitations in the aeroservoelastic codes used to perform turbine simulations. This work utilizes a new aeroservoelastic code developed at Delft University of Technology to model the NREL/Upwind 5 MW wind turbine to investigate the relative advantage of utilizing multiple-device AALC. System identification techniques are used to identify the frequencies and shapes of turbine vibration modes, and these are used with modern control techniques to develop both Single-Input Single-Output (SISO) and Multiple-Input Multiple-Output (MIMO) LQR flap controllers. Comparison of simulation results with these controllers shows that the MIMO controller does yield some improvement over the SISO controller in fatigue load reduction, but additional improvement is possible with further refinement. In addition, a preliminary investigation shows that AALC has the potential to reduce off-axis gearbox loads, leading to reduced gearbox bearing fatigue damage and improved lifetimes.

The optimal operation of an integrated electricity and natural gas system is investigated. The couplings between these two systems are modeled by energy hubs, which serve as interface between the loads and the transmission infrastructures. Previously, we have applied a distributed control scheme to a static three-hub benchmark system. In this paper, we propose an extension of this distributed control scheme for application to energy hubs with dynamics. The dynamics that we consider here are due to storage devices present in the multi-carrier system. We propose a distributed model predictive control approach for improving the operation of the system by taking into account predicted behavior and operational constraints. Simulations in which the proposed scheme is applied to the three-hub benchmark system illustrate the potential of the approach.

The recent progress in the chemical industry is now forcing engineers and physicists to get to deal with control-oriented modeling of material properties on microscopic scale inside reactors in order to build more efficient chemical plants. The control-oriented modeling provides a new way of thinking about the purposes of models in the chemical industry, the relationships between the material properties inside the chemical reactors, and the level of details needed for control designs. The main objective of this thesis is to offer the physics, chemical, and control communities a unified set of rules and conditions for the control-oriented microscopic modeling of fluid systems in the chemical industry. The work presented in this thesis includes not only the physics of fluid systems and first principle models, but also suggests requirements for developing causal input/output structures and spatially distributed control designs. The first part of this thesis concerns the problem of the microscopic modeling of a fluid flow system and the requirements needed to obtain a control-oriented model of the fluid flow system. The derivation of the control-oriented model can be a rather demanding task due to the fact that the dynamics of a particular flow regime have to be determined explicitly. In order to demonstrate the flow control, a single-phase flow system in simple geometry, such as a lid-driven cavity case, is considered. The complexity of the single-phase flow model is illustrated using the Navier-Stokes equations and different discretization methods. The conventional approach to the microscopic fluid flow model involves fine discretization of the microscopic model in order to obtain microstates which can be manipulated and measured. However, the microstates give a very detailed picture of the fluid flow that, in many cases, is not directly measurable. Therefore, different modeling scales have to be considered for designing spatially distributed control strategies for the single-phase flow. In order to demonstrate the applicability of the different modeling scales for the fluid flow system, a macroscopic output regulation is studied on the lid-driven cavity case, which has a broad range of industrial applications. The second part of this thesis deals with the problem of deriving a control-oriented model of a two-phase fluid flow system, which is complementary to the single-phase flow model given in the first part of the thesis. In essence, the control-oriented two-phase flow modeling namely means deriving a simplified model that is available from the first principles and to examine the dominant dynamics. The control-oriented model of the two-phase flow investigates the possibility of identifying different flow regimes inside a bubble column reactor, where the fluid is injected at different locations of the reactor. Besides being useful for control, the control-oriented model of the two-phase flow inside the bubble column also suggest new reactor designs based on the most efficient actuation strategies. Following a wide range of possible actuation structures for the two-phase fluid flow, different spatially distributed control designs for the two-phase flow inside the bubble column reactor are suggested in this thesis. The spatially distributed control strategies can be successfully used to stabilize or destabilize the two-phase flow around a desired two-phase flow regime. In general, the stabilization or destabilization of the fluid flow plays a crucial role in designing efficient and sustainable processes that rely on the manipulation of hydrodynamics. The spatially distributed control designs of the single-phase and two-phase flow systems presented in this thesis suggest more efficient reactor designs and new developments in the process intensification in the chemical industry.

The optimization of combined electricity and natural gas systems is addressed in this paper. The two networks are connected via energy hubs. Using the energy hub concept, the interactions between the different infrastructures can be analyzed. A system consisting of several interconnected hubs forms a distributed power generation structure where each hub is controlled by its respective control agent. Recently, a distributed control method has been applied to such a system. The overall optimization problem including the entire system is decomposed into subproblems according to the control agents. In this paper, a parallel and serial version of that method is discussed. Simulation results are obtained through experiments on a three-hub benchmark system.