Workshops
The ACC will offer workshops addressing current and future topics in automatic control from experts in academia, national laboratories, and industry. The workshops at ACC 2015 will take place prior to the conference on Monday, June 29 and Tuesday, June 30. Please note that workshops are subject to (a) cancellation due to lack of registrants and (b) capacity limits.
Conference registrants can sign up for workshops directly through the registration site.
Workshop Schedule
Monday and Tuesday, June 29 and 30, 2015
2-day workshop (8:30 – 5:30)
Model Predictive Control Workshop
Organizer: James B. Rawlings
Location: Salon 1
Monday, June 29, 2015
Full-day workshops (8:30am – 5:30pm)
Game Theory: Models and Applications to Networked Systems
Organizer: Angelia Nedich
Location: Salon 2
New Advances in Uncertainty Analysis and Estimation
Organizers: Puneet Singla, Raktim Bhattacharya
Location: Salon 3
Nonlinear Optimization: Techniques for Engineering
Organizer: R. Russell Rhinehart
Location: Salon 4
Tuesday, June 30, 2015
Full-day workshops (8:30am – 5:30pm)
Control of Nonlinear Physical Systems
Organizers: Roger Brockett, P.S. Krishnaprasad, Tudor Ratiu, Dmitry Zenkov
Location: Salon 2
Decision Making Algorithms for Unmanned Vehicles
Organizers: Krishnamoorty Kalyanam, Sivakumar Rathinam, Swaroop Darbha
Location: Salon 7
Identification of Linear, Parameter Varying, and Nonlinear Systems: Theory, Computation, and Applications
Organizer: Wallace E. Larimore
Location: Salon 9
Half-day workshops (8:30am – 12:30pm)
Enabling the Grid of the Future
Organizers: Brian Johnson, Srinivasa Salapaka, Blake Lundstorm, Matt Wytock, Zico Kolter, Murti V. Salapaka
Location: Salon 3
Half-day workshops (1:30pm – 5:30pm)
Next Generation Smart Grids: Power Electronics Based Power Systems
Organizer: Qing-Chang Zhong
Location: Salon 3
Robust and Adaptive Control with Aerospace Examples
Organizers: Kevin A. Wise, Eugene Lavretsky
Location: Salon 5
Taxonomies of Interconnected Systems: Asymmetry and Directedness in Multi-Agent Interactions
Organizers: Andrea Gasparri, Ryan K. Williams, Frank L. Lewis
Location: Salon 12
Workshop Descriptions
Monday and Tuesday, June 29 and 30, 2015
Model Predictive Control Workshop
Organizer: James B. Rawlings (University of Wisconsin)
Additional speaker: Thomas A. (Tom) Bagwell (ExxonMobil Research & Engineering)
Model predictive control (MPC) has become the most popular advanced control method in use today. Its main attractive features are (i) optimization of a model forecast over the available actuators (ii) estimation of the state of the system and disturbances from the process measurements, (iii) accounting for the process and actuator constraints, and (iv) accounting for full multivariable interactions. After its introduction in the process industries in the 1970s, MPC has today become a pervasive control technology in many industries, and is now being increasingly deployed for optimization of high-level functions such as minimizing energy consumption and maximizing product quality.
This short course is intended to introduce graduate students and practitioners to the theory and design of MPC systems.
The two days of lectures will cover the following topics.
- Introduction, dynamic modeling, predictive control versus classical PID control.
- Model predictive control: regulation problem, linear quadratic regulator, constraints, dynamic programming, infinite horizon, LQR, constrained regulation.
- State estimation: least-squares estimator, Kalman filter, observability and convergence.
- Putting regulation and estimation together, industrial practice, disturbance models, and offset.
- Nonlinear MPC. introduction, stability, Lyapunov function theory, disturbances and robust stability, nominal stability, suboptimal MPC, inherent robustness of optimal and suboptimal MPC, some examples.
- Other topics: distributed MPC, economic MPC, and hybrid MPC.
In addition, software for implementing linear and nonlinear MPC regulators and moving horizon estimators will be presented. Solving exercises with provided numerical software in open-source Octave (www.octave.org) (or Matlab) and open-source Python/CasADi (www.casadi.org) will be part of the workshop, and registrants should plan on bringing their own laptops to the class.
Additional information can be found here.
Monday, June 29, 2015
Game Theory: Models and Applications to Networked Systems
Organizer: Angelia Nedich (University of Illinois, Urbana-Champaign)
Additional speakers: Tamer Basar (University of Illinois, Urbana-Champaign), Roland Malhame (University of Montreal), Jason Marden (University of Colorado)
The goal of this workshop is to provide an introduction into basic game-theoretic concepts and tools, and to showcase some recent applications of game theory in control of emerging large scale and distributed networked systems. The workshop comprises the following topics.
Foundations of Game-Theoretic Framework for Networks and Control (Basar). With its rich set of conceptual, analytical and algorithmic tools, game theory has emerged as providing a versatile and effective framework for addressing a multitude of issues in networks and control, including resilience, reliability and security in networked (control) systems. This expository talk will introduce the key elements of this modeling paradigm, and discuss various game-theoretic solution concepts, mostly within the framework of nonzero-sum games. Among these are the solution concepts of saddle point (for zero-sum games) and Nash equilibrium as well as Stackelberg equilibrium (for nonzerosum games), for both static and dynamic games, as well as stochastic games. The talk will also cover efficiency (or inefficiency) of these solutions within a non-cooperative mode of decision-making, their sensitivity to imprecision in modeling, and ways of coping with the presence of strategic adversaries. Further, the role of incentive (or disincentive) mechanisms in mitigating or totally eliminating the adverse effects of inefficiency, sensitivity, and adversarial impact will be discussed. The presentation will conclude with some specific applications of the game-theoretic framework in networked control, sensor networks, and cyber-physical systems.
Mean Field Control Theory and its Applications (Malhame). The fundamental intuitions that underline the development of so called Mean Field Games (also known as Mean Field Control Theory) will be presented, as well as some of its foundational results for continuous time systems. The results for both linear and nonlinear continuous time systems will be discussed. Mean Field Control emerges as the natural tool for dealing with the coordination and decentralized control of systems made up of large aggregates of similar weakly interacting elements such as found in the natural world from herds, to fish schools, to beehives, to human societies. Such configurations also occur in manmade constructs such as economic systems and the Internet. We present applications of the linear quadratic versions of the theory, first to a collective navigation problem such as fish schooling; secondly, to a class of control problems in the area of smart grids, whereby large collections of energy storage capable devices such as electric water heaters, are coordinated to mitigate the variability of renewable energy sources.
Games, Information, and Networked Control (Marden). Game theory is a well-established discipline in the social sciences that is primarily used for modeling social behavior. Traditionally, the preferences of the individual agents are modeled as utility functions and the resulting behavior is assumed to be an equilibrium concept associated with these modeled utility functions, e.g., Nash equilibrium. This is in stark contrast to the role of game theory in engineering systems where the goal is to design both the agents utility functions and an adaptation rule such that the resulting global behavior is desirable. The transition of game theory from a modeling tool for social systems to a design tool for engineering systems promotes several new research directions that we will discuss in this talk. In particular, this talk will focus on the following questions: (i) How to design admissible agent utility functions such that the resulting game possesses desirable properties, e.g., the existence and efficiency of pure Nash equilibria? (ii) How to design adaptation rules that lead to desirable system-wide behavior? and (iii) How does the information available to the agents impact achievable performance guarantees in distributed engineering systems?
Games on Time-varying Networks (Nedich). This talk will present some special games arising in networked systems with dynamically changing connectivity structure and with limited access to the whole system information. Some examples of such games that will be discussed include Transferable Utility (TU) games, aggregative games and monotone Nash games on graphs, where the players can use the local neighborhoods to learn/estimate network wide quantities that affect their payoff/ cost functions. We will discuss distributed strategies such as decentralized gradient-play strategies that can result in a Nash equilibrium in the presence of imperfect information such as gradient noise and other forms of uncertainties. Also, the complexity estimates for such strategies will be discussed in terms of their scaling properties with the time and with the number of players.
Additional information can be found here.
New Advances in Uncertainty Analysis and Estimation
Organizers: Puneet Singla (SUNY Buffalo), Raktim Bhattacharya (Texas A & M University)
Both sensor observation data and mathematical models are used to assist in the understanding of physical dynamic systems. However, observational data is often limited in terms of the kind and frequency of observations that can be taken and may only provide access to limited aspects of the system states. Also, any mathematical model used to represent the system dynamics is a reflection of numerous assumptions and simplifications to permit determination of a tractable model. These factors cause overall accuracy to degrade as the model states evolve. The fusion of observational data with state models promises to provide greater understanding of physical phenomenon than either approach alone can achieve. The most critical challenge here is to provide a quantitative assessment of how closely our estimates reflect reality in the presence of model uncertainty, discretization errors as well as measurement errors and uncertainty. The quantitative understanding of uncertainty is essential when predictions are to be used to inform policy making or mitigation solutions where significant resources are at stake.
This workshop will focus on recent development of mathematical and algorithmic fundamentals for uncertainty propagation, forecasting, and model-data fusion for nonlinear systems. The emphasis of this workshop will be on an intuitive understanding of the stochastic processes and practical applications of theory of stochastic processes in estimation and control area. The objectives are to develop a fundamental understanding of stochastic processes and its applications in the area of filtering and control of dynamical systems, to develop an appreciation for the strengths and limitations of state-of-the-art numerical techniques for uncertainty propagation and nonlinear filtering, to reinforce knowledge in stochastic systems with particular emphasis on nonlinear and dynamic problems, and to learn to utilize stochastic system analysis methods as research tools. After the completion of this workshop, audience should be able to apply the discussed methods to real engineering problems with the awareness of potential difficulties that might arise in practice. This workshop would cover topics from basic linear and nonlinear stochastic processes to well-known Kalman filtering methods to recently developed nonlinear estimation methods at a level of detail compatible with the design and implementation of modern control and estimation of dynamical systems. These diverse topics will be covered in an integrated fashion, using a framework derived from stochastic processes, estimation, control, and approximation theory. The reliability and limitations of various methods discussed will be assessed by considering various academic and engineering problems. At the end of this workshop, audience will be able to:
- Understand and use the concept of stochastic processes to model engineering systems.
- Learn to apply linear uncertainty propagation and filtering techniques to engineering problems.
- Understand and derive numerical solution techniques to solve nonlinear uncertainty propagation and filtering problems.
- Get exposed to implementation issues such as computational complexity, non-Gaussian uncertainty, reduction of filter dimension, colored noise, discretization etc
The numerical methods to solve the Kolmogorov equation, Perron-Frobenius operator, generalized Polynomial chaos and stochastic collocation methods will be discussed to determine evolution of state pdf due to probabilistic uncertainty in initial or boundary conditions, model parameters and forcing function. Recent advances in sampling methods like Conjugate Unscented Transformation (CUT) will be presented to compute multi-dimensional expectation integrals. A Bayesian framework is used to assimilate the noisy observation data from various sources with uncertain model forecasts to reduce the uncertainty associated with model-state estimates. By accurately characterizing the uncertainty associated with both process and measurement models, this workshop offers systematic design of low-complexity model-data fusion or filtering algorithms with significant improvement in nominal performance and computational effort. Various academic and engineering problems where traditional methods either fail or perform very poorly, will be considered to demonstrate the reliability and limitations of the newly established methods.
Additional information can be found here
and https://stochasticsystems.wordpress.com/.
Nonlinear Optimization: Techniques for Engineering
Organizer: R. Russell Rhinehart (Oklahoma State University)
This full-day workshop will be a practical guide for those using multivariable, constraint handling, nonlinear optimization. Although theoretical analysis behind techniques will be revealed, the takeaway will be participants’ ability to:
- Define the objective function (cost function)
- Incorporate hard constraints in the search
- Convert constraints to appropriately weighted penalties
- Choose an appropriate optimizer
- Choose appropriate convergence criteria and thresholds
- Choose initialization and number of trials to be confident in finding the global optimum
The short-course will cover common gradient-based optimization techniques (Incremental Steepest Descent, Cauchy, Newton, Levenberg-Marquardt) and direct-search techniques (Heuristic, Hook-Jeeves, Particle Swarm, and Leapfrogging) that represent the fundamentals of most approaches. This workshop is based on a popular interdisciplinary graduate engineering course. Participants will receive course notes of approximately 200-pages, and software to provide exercises and access to code. Exercises and code can be implemented in any environment, but Excel/VBA will used as in-workshop examples and exercises. Participants are invited to bring a computer with Excel version 2010 or higher for in-class exploration. There are currently more than 80 functions and over 20 optimization algorithms in the simulator.
The major challenges in optimization are often not 1) the mathematics of the algorithm, but the clear and complete statement of the 2) objective function, 3) constraints, 4) decision variables, 5) models, 6) convergence criterion, 7) initialization, 8) starting points, and 9) optimization algorithm. This short-course addresses all nine of the elements.
Additional information can be found here.
Tuesday, June 30, 2015
Control of Nonlinear Physical Systems
Organizers: Roger Brockett (Harvard University), P.S. Krishnaprasad (University of Maryland), Tudor Ratiu (Ecole Polytechnique Federale de Lausanne), Dmitry Zenkov (North Carolina State University)
Additional speakers: John Baillieul (Boston University), Roger Brockett (Harvard University), Manuel de Leon (Instituto de Ciencias Matemáticas), Hermann Flaschka (University of Arizona), Jessy Grizzle (University of Michigan), Harris McClamroch (University of Michigan), Alberto Rojo (Oakland University), Michael Weinstein (Columbia University)
Overview:
The workshop will provide a comprehensive view of the subject of nonlinear control of physical systems. As geometry and calculus of variations are of fundamental importance for physics and nonlinear control, the workshop will promote the use of these techniques in both theoretical engineering and applications. It will focus on the structure of configuration/state spaces, symmetry and reduction in dynamics and control, exactly-solvable nonlinear control problems, control and stabilization of underactuated systems with constraints, and numerical techniques for nonlinear control, including structure-preserving geometric integrators. Examples will be drawn from recent work, chosen to illustrate the broad range of applicability these methods have in engineering. Professor Anthony M. Bloch is a leading researcher and widely recognized contributor to the field of nonlinear dynamics and control. His 60th birthday is celebrated with this workshop devoted to the exploration of nonlinear control methods as applied to physical systems, both classical and quantum, the areas in which Professor Bloch made fundamental contributions. He is a fellow of the IEEE, AMS, and SIAM, Editor in Chief of SIAM Journal on Control and Optimization, and current or former member of editorial boards of many excellent journals.
Target Audience:
Younger professionals and graduate students with an interest in developing their expertise in the area of physical and cyber-physical systems as well research teams focused on specific projects involving the formulation and solution of practical problems that can benefit from deeper appreciation of nonlinearity.
Morning and Early Afternoon:
Invited lectures
Late Afternoon:
Discussions and short presentation by selected junior participants
Additional information at: http://www4.ncsu.edu/~dvzenkov/conferences/acc2015.html
Decision Making Algorithms for Unmanned Vehicles
Organizers: Krishnamoorty Kalyanam (Air Force Research Laboratory), Sivakumar Rathinam (Texas A & M University), Swaroop Darbha (Texas A & M University)
Additional speakers: Meir Pachter (Air Force Institute of Technology), Pramod Khargonekar (University of Florida)
Automation tools that account for strategic and tactical decision making involving UVs has received significant attention by the research community over the last decade. This workshop will outline the challenges, disseminate the most recent developments, and present some open problems in this research area. The interest in UV autonomy is expected to increase given the usefulness of UVs in civilian applications such as forest fire monitoring, remote sensing in agricultural and weather monitoring applications. Topics include:
Combinatorial Motion Planning for a Collection of Unmanned Vehicles. (Darbha) This presentation will focus on the combinatorial motion planning problem of allocating tours for a collection of vehicles. The motion of the vehicles satisfies a non-holonomic constraint, i:e:, the yaw rate of the vehicle is bounded. Each target is to be visited by one and only one vehicle. Given a set of targets and the yaw rate constraints on the vehicles, the problem is to assign each vehicle, a sequence of targets to visit, and to find a feasible path for each vehicle that passes through the assigned targets with a requirement that the vehicle returns to its initial position. The heading angle at each target location may not be specified. The objective function is to minimize the sum of the distances traveled by all the vehicles. This presentation will present an overview of the available methods and the latest developments for both the single and the multiple vehicle case.
Approximation Algorithms for Routing a Team of Heterogeneous Unmanned Vehicles (Rathinam) Heterogeneity of UVs is an important issue that complicates the motion planning of UVs with motion constraints. It can arise in different ways: (1) Vehicles may be structurally heterogeneous, i:e:, they can be of different make and hence, the cost of traveling from one target to another may depend on the employed vehicle, or (2) Vehicles may be functionally heterogeneous, i:e:, vehicles are identical structurally, but each vehicle may have a different suite of sensors which imposes some restrictions on the targets that may be visited by each vehicle and vice-versa. The underlying combinatorial problem is difficult because it embeds and couples two difficult combinatorial problems: (1) partitioning the set of targets that may be visited by each vehicle, and (2) sequencing the set of targets that may be visited by each vehicle. In this presentation, we present novel approximation algorithms for constructing tours for a structural and functional heterogeneous collection of UVs. These algorithms will be based on rounding linear programs and primal-dual methods, and will disseminate the latest developments in this area.
Informational Issues in Dynamic Games (Pachter) The state of affairs concerning dynamic games with incomplete information is not satisfactory. Upon reviewing the literature it quickly becomes apparent that there is an acute need to clarify critical conceptual issues. In this respect, the situation is not much different now than it was in 1971 when, Witsenhausen, in his IEEE Proceedings paper, made a similar observation. In this talk, a careful analysis of conceptual issues in dynamic games with incomplete information and decentralized optimal control is undertaken. Informational issues arising in stochastic dynamic games involving UVs and adversaries are addressed. The emphasis is on conceptual issues and gaining insight - we shall therefore exclusively focus on Linear-Quadratic Dynamic Games (LQDGs), which are more readily amenable to analysis. At the same time, LQDGs stand out as far as applications of the theory of dynamic games are considered. In particular, the talk will highlight the critical role of information in dynamic games that arise in UV decision making. Example scenarios such as pursuit of an intruder by UVs on a road network will be discussed.
Approximate Dynamic Programming and its Application to UAV Perimeter Patrol (Kalyanam) This talk addresses the following problem: an Unmanned Aerial Vehicle (UAV) and a human operator cooperatively perform the task of perimeter surveillance. Alert stations consisting of Unattended Ground Sensors (UGSs) are located at key locations along the perimeter. Upon detection of an incursion in its sector, an alert is flagged by the UGS. A camera equipped UAV is on continuous patrol along the perimeter and is tasked with inspecting UGSs with alerts. Naturally, the longer a UAV dwells (loiters) at an alert site, the more information it gathers and transmits to the operator. The objective here is to maximize the information gained, and, at the same time, reduce the expected response time to an alert. To determine the optimal patrol policy, one has to solve a Markov decision problem, whose large size renders exact dynamic programming methods intractable. So, we explore a state aggregation based approximate linear programming method to construct provably good sub-optimal policies instead. As a general result, it is shown that this approximate value function is independent of the non-negative cost function (or state dependent weights; as it is referred to in the literature) and moreover, this is the least upper bound that one can obtain, given the partitions. Furthermore, it is shown that the restricted system of linear inequalities also embeds a family of Markov chains of lower dimension, one of which can be used to construct a tight lower bound on the optimal value function. Finally, numerical results supporting the approximate method are presented.
Optimal Minimax Pursuit Evasion on a Manhattan Grid (Khargonekar) This presentation focusses on the problem of finding an optimal control for a pursuer searching for a slow moving evader on a road network. The pursuer does not have the on-board capability to detect the evader and relies instead on Unattended Ground Sensors (UGSs) to locate the evader. We assume that all the intersections in the road network have been instrumented with UGSs. When an evader passes by an UGS location, it triggers the UGS and this time-stamped information is stored by the UGS. When the pursuer arrives at an UGS location, the UGS informs the pursuer if and when the evader passed by. When the evader and the pursuer arrive at an UGS location simultaneously, the UGS is triggered and this information is instantly relayed to the pursuer, thereby enabling “capture”. In this presentation, we will outline the optimal strategy, provide a tight bound on the optimal number of steps to capture and showcase the main results.
Additional information can be found at https://sites.google.com/site/srathinam/workshop.
Identification of Linear, Parameter Varying, and Nonlinear Systems: Theory, Computation, and Applications
Organizer: Wallace E. Larimore (Adaptics, Inc.)
In this workshop, the powerful subspace identification method (SIM) is described for the well understood case of linear time-invariant (LTI) systems. Recent extensions are then developed to linear parametervarying (LPV), Quasi-LPV, and general nonlinear (NL) systems such as polynomial systems. The presentation, following the extended tutorial paper (Larimore, ACC2013), includes detailed conceptual development of the theory and computational methods with references to the research literature for those interested. Numerous applications including aircraft wing flutter (LPV), chemical process control (LTI), automotive engine (Quasi-LPV, NL) modeling, and the Lorenz attractor (NL) are discussed. An emphasis is placed on conceptual understanding of the subspace identification method to allow effective application to system modeling, control, and fault diagnosis.
Over the past decade, major advances have been made in system identification for the LTI cases of no feedback (Larimore, ACC1999) and unknown feedback (Larimore, 2004; Chiuso, TAC2010). However, for LPV and NL systems limitations remain including, for subspace methods the required computation grows exponentially with the number of system inputs, outputs, and states, and for maximum likelihood methods iterative nonlinear parameter optimization may not convergence, leading often to infeasible computation.
The workshop presents a first principles statistical approach using the fundamental canonical variate analysis (CVA) method for subspace identification of linear time-invariant (LTI) systems, with detailed extensions to linear parameter-varying (LPV) and nonlinear systems. The LTI case includes basic concepts of reduced rank modeling of ill-conditioned data to obtain the most appropriate statistical model structure and order using optimal maximum likelihood methods. The fundamental statistical approach gives expressions of the multistep-ahead likelihood function for subspace identification of LTI systems. This leads to direct estimation of parameters using singular value decomposition type methods that avoid iterative nonlinear parameter optimization. The result is statistically optimal maximum likelihood parameter estimates and likelihood ratio tests of hypotheses. The parameter estimates have optimal Cramer-Rao lower bound accuracy, and the likelihood ratio hypothesis tests on model structure, model change, and process faults produce optimal decisions. Comparisons made between system identification methods including subspace, prediction error, and maximum likelihood, and show considerably less computation and higher accuracy.
The LTI subspace methods are extended to LPV systems that are expressible in the LTI form where the constant LTI parameters are multiplied by parameter-varying scheduling functions depending on the system operating point. For example, this allows for the identification of constant underlying structural stiffness parameters while wing flutter dynamics vary with scheduling functions of speed and altitude operating point variables. This is further extended to Quasi-LPV systems where the scheduling functions may be functions of the inputs and/or outputs of the system. Quasi-LPV systems include bilinear and general polynomial systems that are universal approximators. The developed subspace identification method for parametervarying systems avoids the exponential growth in computations characteristic of previous SIM methods. Applications are discussed to monitoring and fault detection in closed-loop chemical processes, identification of vibrating structures under feedback, adaptive control of aircraft wing flutter, identification of the chaotic Lorenz attractor, and identification and monitoring of Quasi-LPV automotive engines.
Additional information can be found here.
Enabling the Grid of the Future
Organizers: Brian Johnson (National Renewable Energy Laboratory), Srinivasa Salapaka (University of Illinois, Urbana-Champaign), Blake Lundstorm (National Renewable Energy Laboratory), Matt Wytock (Carnegie Mellon University), Zico Kolter (Carnegie Mellon University), Murti V. Salapaka (University of Minnesota)
In this workshop, the realization of systems that enable integration of distributed generation, while incorporating objectives related to economics and reliability, will be targeted. An attendee of the workshop will, by the end of the workshop, attain the fundamentals of how to design, implement, and analyze a power electronics based system that enables distributed generation, prioritization, and integration of heterogeneous power sources. The key modeling and control paradigms will be described and recent research results will also be elucidated. Key insights into how to monitor such systems using machine learning tools will be emphasized. Finally, directions on how modern control tools can be brought to bear will be proposed. Topics include:
Grid of the future (S. Salapaka, M. Salapaka) Here, the present state of the grid and the possibly disruptive changes in the near future that are impending will be highlighted. This discussion will set the stage for why new research and education in controls is needed in this area and the niche areas where controls can play a future role will be highlighted. The overall architecture of power flow of the present grid architecture will be summarized.
Control of energy delivery using power electronics (Johnson, Lundstrom) The latest advances in digital control and power electronics circuits enable the development of flexible and high-bandwidth energy delivery systems. In any type of modular power electronics system, the instance of a single power converter forms the basic building block. Accordingly, this portion of the tutorial will provide a brief overview of the physical construction and characteristics of a power converter, pulse width modulation techniques, development of average models, and design of closed-loop controllers for power converters. The talk will be focused on ac systems which utilize dcac inverters. Using the presented material, the audience member will have the requisite knowledge to model a power electronics inverter, design its controller, and have a working understanding of the implementation issues at the single power converter level. A special emphasis will be placed on the typical assumptions which facilitate analytical design tools and their underlying physical interpretations. Furthermore, recent results which utilize the modern control framework for the formulation of optimal voltage controllers will be introduced. It will be shown that the optimal controller exhibits an inner-outer structure and corroborates the longstanding observation that nested loop controllers have superior performance. The modeling and control design framework will be presented with a level of generality such that any higher-level controller may be augmented to the primary-level controller which is presented.
System with multiple generation sources (Johnson, Lundstrom, S. Salapaka, M. Salapaka) Current electrical grids are semi-centralized with a top-down power flow structure, where large generating stations produce electrical power, high-voltage transmission lines carry power from distant sources to demand centers, and distribution lines connect individual customers. This grid architecture inherently does not support small power sources and does not exploit the proximity of the power sources to the loads. It relies heavily on the a priori estimation of loads and is not designed to accommodate large uncertainties in power generation or consumption. An alternate architecture is to implement plug and play, bottom-up grids that use local power generation for local loads and interconnect to exchange power. This architecture and its use of multiple modular renewable energy sources, such as wind and photovoltaics, with special emphasis on management of uncertainties of these resources will be presented. The design objectives and challenges in realizing such a system will be described. The main objectives include voltage regulation and achieving sharing/prioritization between multiple power sources. The challenges include complexity in terms of architectures (sharing on dc side vs ac side vs both) and control of power-electronics to enable multiple objectives stated above, preserve power quality, and ensure seamless transitions between grid tied and off-grid configurations. Conventional as well as new approaches based on modern robust control framework and nonlinear oscillator based strategies to address these challenges will be explained.
Monitoring and prediction with machine learning (Kolter, Wytock) Here, real-time paradigms for monitoring and prediction will be highlighted, focusing on fast approaches well-suited for practical applications related to power networks. Modern methods for analyzing large-scale time series data are emphasized and the formulation of problems in a machine learning framework will be discussed.
Additional information can be found here.
Next Generation Smart Grids: Power Electronics Based Power Systems
Organizer: Qing-Chang Zhong (Illinois Institute of Technology)
Power systems are going through a paradigm change from centralized generation to distributed generation and further onto smart grids. In order to make power systems more secure, more efficient, more resilient to threats and friendlier to the environment, a huge number of heterogeneous players, including renewable energy sources, electric vehicles, and storage systems etc. on the supply side and different types of smart loads on the demand side, are being connected to power systems. Because of the heterogeneous nature and the huge number of players involved, it is a great challenge for control and systems theorists to find a control architecture so that all heterogeneous players could work together to maintain system stability and achieve desired performance.
In this workshop, the fundamental challenge behind the scene during the paradigm change is identified: that is future power systems will be power electronics based, instead of electrical machines based, with a huge number of heterogeneous players. This makes it less of a power problem but more of a systems problem. Moreover, an autonomous scalable distributed control architecture is presented from the systems perspective. All the heterogeneous players, including new add-ons of generation, such as wind farms, solar farms, EVs, energy storage systems, and the majority of loads, can be controlled to behave like virtual synchronous machines so that all behave homogeneously, in terms of the underlying mathematical models. This unifies the interface of all these players with the grid and facilitates the reduction of large-scale power systems into small-scale models and the analysis of power systems. All the distributed players (agents) communicate with each other through the dynamics of power systems, instead of an extra communication network, to realize the same goal with independent individual actions. Because the low-level control does no longer require the support of an extra communication network, this paradigm is distinct from the current paradigm of smart grids and hence sets the architecture for the next-generation smart grids. It is able to considerably enhance the operability, stability, scalability, reliability and security of next-generation smart grids.
Two technical routes to implement the architecture will be presented. One is based on the synchronverter technology that takes into account the internal dynamics of synchronous machines and the other is based on the robust droop control strategy that mimics the external function of synchronous machines. Both technical routes embed the synchronisation function into the controller of the power electronic converters and hence the dedicated synchronization unit, e.g. PLL, that is deemed to be a must-have component for grid-tied power electronic converters can be removed.
Additional information can be found here.
Robust and Adaptive Control with Aerospace Examples
Organizers: Kevin A. Wise, Eugene Lavretsky (The Boeing Company)
This is a two-part workshop that covers robust control, used to form a baseline control, and adaptive control, used to extend the baseline control system’s robustness and performance under uncertainties. Both state feedback and output feedback architectures are presented. Robust and Adaptive Control with Aerospace Applications, Lavretsky, Wise, SpringerVerlag, 2013, is included.
Part I begins with an introduction to challenges in control design, analysis, and simulation of manned and unmanned aircraft. General aviation background and current trends that lead to the need for more advanced control are discussed. Also presented is a brief survey of control-theoretic methods for existing and future aircraft. The theoretical portion of Part I starts with the introduction of robust and optimal linear control methods for linear systems. Command tracking using linear quadratic regulators (LQR) with integral action is presented. This part also covers two output feedback design methods, such as projective control and linear quadratic Gaussian control with Loop Transfer Recovery (LQG/LTR). These algorithms are employed to develop a baseline control architecture. New developments in observer-based architectures, called observer-based loop transfer recovery (OB-LTR), asymptotically achieves positive real system behavior at certain loop break points during recovery. This workshop will present design insights into using this method for flight control design problems, including systems using acceleration feedbacks that are nonminimum phase. During the design process the observer is artificially squared-up. This adds fictitious inputs to make the number of controls equal the number of measurements, and makes the observer design model minimum phase. This step is central to achieving the positive real behavior during recovery. To place the zeros in a desired location during plant squaring, an LQR or pole-placement algorithm can be used. Tutorial design examples will be covered to highlight this new design approach.
Part II begins with self-contained material on the design and analysis of adaptive state feedback controllers for linear and nonlinear uncertain dynamical systems in continuous-time domain. An overview of Lyapunov stability theory is given, followed by theoretical fundamentals for MRAC systems. Next, approximation properties of artificial neural networks and their applications to the design of direct adaptive systems are introduced, and several approximation-based MRAC methods are discussed. The part proceeds with the development of state feedback adaptive augmentation architectures for robust baseline linear controllers, followed by extensions and modifications to achieve transient performance in adaptive systems, as well as to accommodate output feedback constraints. In this part, we also present adaptive augmentation design methods to combine robust baseline controllers with adaptive feedback, focused on using the OB-LTR architecture.
Additional information can be found here.
Taxonomies of Interconnected Systems: Asymmetry and Directedness in Multi-Agent Interactions
Organizers: Andrea Gasparri (Universita degli studi \"Roma Tre"), Ryan K. Williams (University of Southern California), Frank L. Lewis (University of Texas Arlington)
Additional speakers: Fabio Pasqualetti (University of California, Riverside), Guodong Shi (Australian National University), Wei Ren (University of California, Riverside), Jorge Cortes (University of California, San Diego), Amir G. Aghdam (Concordia University), Magnus Egerstedt (Georgia Institute of Technology)
Interconnected systems have become the recent focus of intense investigation, particularly in the context of autonomous collaboration (such as in multi-robot or sensor systems), according fundamental advantages in adaptability, scalability, and efficiency compared to classical single-agent solutions. As recent work has demonstrated, investigations are far-reaching across various disciplines, ranging from sampling, tracking, and coverage, mobility and topology control, to general agent agreement problems. In this workshop, we will focus on the peculiarities arising from directedness and asymmetry in pairwise agent-to-agent interactions. Topics include:
Controllability Properties of Complex Networks (Pasqualetti) Network science is an interdisciplinary research area that focuses on dynamic processes over networks. Network science problems arise in several domains including sensor and actuator networks, robotics, social networks, and biological systems. The main objectives are to analyze, predict, and control complex behaviors over networks. In this talk I will discuss controllability properties of complex networks, with respect to their topology and weights, and to the number and location of the control nodes. I will present metrics to quantify network controllability, bounds on the number of control nodes to ensure a certain controllability degree, and a distributed control strategy with performance guarantees. I will show that isotropic networks are difficult to control, as the control energy grows exponentially with the network cardinality when the number of control nodes remains constant. Conversely, sufficiently certain anisotropic networks are easy to control, as the control energy is bounded independently of the network cardinality and number of control nodes.
Influence of Directed or Bidirectional Communications in Distributed Consensus and Gossip Algorithms (Shi) In networked control systems, the information flow among the nodes can be directed or bidirectional. In many occasions we can treat them consistently, while sometimes they can behave in quite different manners. In this talk, I will discuss how directed or bidirectional communications influence distributed consensus and gossip algorithms. Both classical networks and quantum networks will be investigated, and some basic consistency and difference between the two types of networks will be addressed. I will show that bidirectional interactions indeed provide a lot convenience for most times, but, at least sometimes, essentially directed node interactions can also be irreplaceable.
Distributed Average Tracking in Multi-agent Networks (Ren) In this talk, we introduce a distributed average tracking problem and present distributed discontinuous control algorithms to solve the problem. The idea of distributed average tracking is that multiple agents track the average of multiple time-varying reference signals in a distributed manner based only on local information and local communication with adjacent neighbors. We study the cases where the time-varying reference signals have bounded derivatives and accelerations. We also use the distributed average tracking idea to solve a continuous-time distributed convex optimization problem. Tools from non-smooth analysis are used to analyze the stability of the systems. Simulation examples are presented to show the validity of the theoretical results.
Cooperative Control Synchronization: Optimal Design and Games on Communication Graphs (Lewis) In this talk we present design methods for cooperative controllers for distributed systems. The developments are for general directed graph communication structures, for both continuous- time and discrete-time agent dynamics. Cooperative control design is complicated by the fact that the graph topology properties limit what can be achieved by the local controller design. Thus, local controller designs may work properly on some communication graph topologies yet fail on other topologies. Our objective is to provide local agent feedback design methods that are independent of the graph topology and so function on a wide range of graph structures. An optimal design method for local feedback controllers is given that decouples the control design from the graph structural properties. In the case of continuous-time systems, the optimal design method guarantees synchronization on any graph with suitable connectedness properties. In the case of discrete-time systems, a condition for synchronization is that the Mahler measure of unstable eigenvalues of the local systems be restricted by the condition number of the graph. Thus, graphs with better topologies can tolerate a higher degree of inherent instability in the individual node dynamics. A theory of duality between controllers and observers on communication graphs is given, including methods for cooperative output feedback control based on cooperative regulator designs.
In Part 2 of the talk, we discuss graphical games. Standard differential multi-agent game theory has a centralized dynamics affected by the control policies of multiple agent players. We give a new formulation for games on communication graphs. Standard definitions of Nash equilibrium are not useful for graphical games since, though in Nash equilibrium, all agents may not achieve synchronization. A strengthened definition of Interactive Nash equilibrium is given that guarantees that all agents are participants in the same game, and that all agents achieve synchronization while optimizing their own value functions.
Distributed Event-triggered Coordination for Average Consensus on Weight-balanced Directed Graphs (Cortes) Given the widespread applications of average consensus in distributed control and estimation, it is important to synthesize efficient algorithmic solutions over directed interaction topologies, where information among agents travels directionally. In this talk we look at the problem of synthesizing event-triggered communication and control coordination policies that allow a multi-agent system to achieve average consensus. The communication topology is described by a weight-balanced, strongly connected directed graph. Our proposed event-triggered communication and control strategy does not rely on individual agents having continuous or periodic access to information about the state of its neighbors and does not require agents to have a priori knowledge of any global parameter. We characterize the asymptotic convergence properties of the resulting executions and rule out the presence of Zeno behavior.
Distributed Connectivity Assessment of Underwater Sensors Network (Aghdam) In this presentation, the problem of distributed connectivity assessment for a network of underwater sensors is presented. Motivated by a sufficient condition for asymptotic almost-sure consensus in a network represented by a random directed graph (digraph), vertex connectivity of the expected communication graph is used as a measure of the connectivity of the underwater sensor network. A distributed update scheme is proposed in which the sensors update their perception of the expected communication graph. A learning algorithm is employed by each sensor to update its belief of the probabilities of different graph edges using the broadcast messages it receives. Each sensor uses a polynomial-time algorithm to estimate the degree of vertex connectivity of the expected graph based on its perception of the network graph. The proposed algorithms can also handle changes in the topology of the network such as node addition, node deletion, and time-varying probabilities. A weighted vertex connectivity degree is also proposed which takes the randomness of the network into consideration in the connectivity measure. The results are subsequently used to evaluate the importance of each node in the overall connectivity of the network. The performance of the algorithms is validated by simulation.
Network Symmetry is an Obstruction to Controllability (Egerstedt) In networks of autonomous agents that coordinate their operations based on local interaction rules, one key question is how easy/hard it is for a human operator to interact with these networks. In particular, if one where to view the human interactions as a control input, injected at a subset of the nodes in the network, this question can be cast as a controllability question. And, it turns out that symmetry is an obstruction to controllability, i.e., if the network is organized in a symmetric manner it is harder to control, which goes against our design intuition which tends to favor symmetry. This talk will explore this issue as well as discuss recent results on the controllability of networked control systems.
Additional information can be found at http://gasparri.dia.uniroma3.it/ws/acc15/index.html.
|
