Distinguished Lecturer Program

This is a talk focused on the criticality of thermal systems in almost every domain of energy conversion. Thermal systems are critically important to nearly all domains of energy conversion, and controls are vital to extracting maximal efficiency from the overall system. Understanding the dynamics of transient thermal systems is the first step towards effective control design. While a great deal of understanding of steady-state performance of an overall system already exists, the combined performance of coupled and interconnected systems during transients is still not well described or understood. This becomes more important with increased system complexity or increased transient relevance. Continued improvement in control-oriented modeling will be very valuable in terms of accuracy, speed, etc. With energy as a crucial theme for a sustainable future, it is clear that the Mechanical Engineering community must play a key leadership role in achieving this potential, since the thermal energy domain is one with which we are most familiar.

Modern aircraft are highly complex systems.  This talk will present a particular hierarchical approach to energy and power flow in air vehicles that accommodates multiple power modes. These modes include thermal, fluid, electrical, or mechanical since these are all available in larger aircraft.  In particular, with the current drive towards increased electrification, the management of power onboard aircraft has become an enabler or a bottleneck depending on the point of view.  A key challenge in working across various modes of power flow is the widely varying time scales.  The hierarchy allows for systems operating on different time scales to be coordinated in a controllable manner. It also allows for different dynamic decision-making tools to be used at different levels of the hierarchy based on the needs of the physical systems under control. Additional advantages include the modularity and scalability inherent in the hierarchy.  Additional modules can be added or removed without changing the basic approach. In addition to the hierarchical control, a particularly useful graph-based approach will be introduced for the purpose of modeling the system interactions. The graph approach, like the hierarchy, has benefits of modularity and scalability along with being an efficient framework for representing systems of different time scales.  Recent results will be presented representing both generic interconnected complex systems as well as specific examples and resulting benefits.

High precision motion control is essential for a wide variety of modern applications. The key to high precision is the incorporation of feedforward information along with typical feedback algorithms. Iterative Learning Control is a very popular method to determine signal-based feedforward control. This talk will discuss recent developments in improving the performance of ILC schemes and their applications to manufacturing applications. In particular, we will motivate the use of ILC schemes with precision manufacturing applications particularly to the nanoscale. We begin by detailing components of a heterogeneous integration approach to the manufacturing of novel electronic and photonic devices by fluidic and ionic transport. This is part of an interdisciplinary research effort involving Materials Science, Physics, Chemistry, Manufacturing, and Controls. The particular process to be detailed is a printing process, termed electro-hydrodynamic Jet (or e-Jet) printing, that is currently superior to most other printing approaches in terms of resolution.

After the demonstration of manufacturing processes, a brief introduction to Iterative Learning Control (ILC) will be given. ILC is a novel adaptive technique that allows us to learn repeated trajectories and maximize precision in the automation machinery used for fabrication. After an overview, the rest of the talk will discuss recent developments in ILC for both single-axis and multi-axis systems. We demonstrate the benefits in performance with numerical and experimental results

Linear systems with multiplicative, time-varying noise exhibit varied and rich phenomenology such as heavy tails and dramatic differences between different notions of convergence. We study such systems in a framework similar to that used in robust control where the stochastic parameters are viewed as a "structured uncertainty". In particular, a purely input-output approach is developed to characterize mean-square stability. This approach clarifies earlier results in this area and also easily produces new ones in the case of correlated uncertainties. Applications of this framework to networked dynamical systems with link failures and stochastic topologies will be illustrated. In addition, an application to a model of the Cochlea will be described which potentially explains otoacoustic emissions as an instability mechanism. Finally, we illustrate some interesting connections of this work with the phenomenon of Anderson Localization which is a canonical problem in the statistical physics of disordered media. 

The question of how difficult or easy it is to control a certain network of interconnected dynamical agents is fundamental to understanding engineered or naturally occurring networks, such as vehicular formations or power grids amongst many others. I will argue that standard notions of stability and controllability as binary properties (e.g. a system is either stable or not), convergence rates, or even reachability analysis may fail to predict the behavior of large networks. These apparent difficulties motivate a notion of network controllability based on hard limits on performance in optimal control problems with structural constraints. While such problems are known to be generally intractable, I will show certain examples from vehicular platoons and power grids where informative and simple answers are possible in the asymptotic limit of large system size. This analysis gives asymptotic bounds on network performance and shows its dependence on both the complexity of individual node dynamics,  as well as network connectivity. Some interesting connections between these results and the statistical mechanics of disordered media will be highlighted. 

In control theory, complex models of physical processes, such as systems of differential equations, are analyzed or controlled from simple specifications, such as stability and set invariance. In formal methods, rich specifications, such as formulae of temporal logics, are checked against simple models of software programs and digital circuits, such as finite transition systems. With the development and integration of cyber-physical and safety-critical systems, there is an increasing need for computational tools for verification and control of complex systems from rich, temporal logic specifications.

In this talk, I will discuss a set of approaches to the formal synthesis of control strategies for dynamical systems from temporal logic specifications. I will first show how automata games for finite systems can be extended to obtain conservative control strategies for low dimensional linear and multilinear dynamical systems. I will then present several methods to reduce conservativeness and improve the scalability of the control synthesis algorithms for more general classes of dynamics. I will illustrate the usefulness of these approaches with examples from robotics and traffic control.

This talk takes into account the current status of automation science and technology development as well as the existing automation undergraduate programs in many Chinese universities at the moment; while lending from the successes in automation science and technology developmental history combined with future demands for automation systems to aid in the economic development and national security of China with emerging technologies including mobile internet, cloud computing and big data driven artificial intelligence; taking manufacturing systems, important vehicles and cyber-physical and human system as research objects, this talk proposes that the future of automation systems be directed towards transforming into intelligent autonomous control system, intelligent optimal decision-making system and integrated system of intelligent optimal decision-making and control. With a research focus geared towards practical application, new algorithms of modeling, control and optimization for the prospective functions of the developing automation systems with subsequent design of methods and implementation techniques for new automation systems are taken as development directions of automation science and technology. Finally, this talk proposes that the future direction for development in the field of automation science and technology should be based on the current challenges and requirements that have emerged in new areas of application.

We are at the cusp of a revolution: we can now create machines that adapt their behavior based on their environment and the results of their actions. The enablers for this revolution are sensing, communication, and computation technologies, and the feedback control algorithms that rule the machines.  In this talk I will discuss the skill set necessary to thrive in this feedback-rich world. 

Hybrid systems, i.e., heterogeneous systems that include discrete and continuous-time subsystems, have been used to model control applications e.g. in automotive control, air traffic management systems, smart grids and intelligent manufacturing. Failure in this kind of applications can cause irreparable damage to the physical controlled systems and to the people who depend on it, or may cause large direct and indirect economic losses. Therefore, security for hybrid systems represent a significant concern.

In this respect, observability and diagnosability play an important role since they are essential in characterizing the possibility of identifying the system’s hybrid state, and in particular, the occurrence of specific states that may correspond to malfunctioning due to a fault or an adversarial attack.

In this talk, I review and place in context how the continuous and the discrete dynamics, as well as their interactions, intervene in the observability and diagnosability properties of a general class of hybrid systems. I also illustrate under which conditions the hybrid system’s state can be correctly estimated even when the system is under attack. An example related to network topology changes due to faults or attacks will illustrate the results.

Freight transportation is of outmost importance for the development of our society and economy. At the same time, transporting goods on roads accounts for a significant amount of all energy consumption and greenhouse gas emissions. Despite this influence, road transportation is mainly done today by individual long-haulage trucks with no real-time coordination or global optimization. In this lecture, we will discuss how modern information and communication technology supports a cyber-physical transportation system architecture with an integrated logistic system coordinating fleets of trucks traveling together in vehicle platoons. From the reduced air drag, platooning trucks traveling close together can save more than 10% of their fuel consumption. Control and estimation challenges and solutions on various level of this transportation system will be presented. It will be argued that a system architecture utilizing vehicle-to-vehicle and vehicle-to-infrastructure communication enables safe and optimal control of individual trucks as well as optimized vehicle fleet collaborations. Empirical evidence will be presented for why large-scale fleet coordination is mainly a scheduling (not a routing) problem. Incentives for cooperation and pricing of transport services will also be discussed. Several experiments done on European highways will illustrate achievable system performance and potential obstacles to be over

Cyber-attacks on critical infrastructures are of growing societal concern. Several malicious attacks have been reported over the last few years and in many cases they have targeted control systems. The increasing use of off-the-shelf software and hardware components and open communication networks makes networked control systems vulnerable to cyber-attacks. As the cyber and physical components of these systems are tightly interconnected, traditional IT security focusing on the cyber part does not provide appropriate solutions. In this talk, we will discuss how to model, analyze and design cyber-secure networked control systems. We will introduce an adversary modeling framework and use it for quantifying cyber-security of control systems by means of constrained optimization problems. An attack space defined by the adversary's model knowledge, disclosure, and disruption resources is presented. It is shown that attack scenarios corresponding to denial-of-service, replay, zero-dynamics, and bias injection attacks can be analyzed using this framework. Applications to power networks and process industry will be used to illustrate the attack scenarios, their consequences, and potential countermeasures.

There is a growing deployment of wireless networks in industrial control systems. Lower installation costs and efficient system reconfigurations for wireless devices have a major influence on the future application of distributed control. Traditional sampled-data control is based on periodic sensing and actuation rather than the acting when the system needs attention.  Event-based control instead is reactive and generates sensor  sampling and control actuation when the plant needs it. In this  talk, we will discuss how to design event-based control systems.  It will be shown how wireless access scheme for can influence  the closed-loop performance of the networked control system.  It will be argued that the underlying scheduling control problem  has a non-classical information structure. Appropriate models  for medium access control protocols will be introduced.  It will be shown how these protocols can be tuned for various  wireless control applications. The talk will be illustrated by  several examples from ongoing projects with Swedish industry.  The presentation is based on joint work with several collaborators.

In this talk we discuss the problem of constructing decentralized control systems, which is an outstanding problem in control theory. For centralized control systems, there are many effective algorithms for computing controllers, and this is possible for a wide class of systems including deterministic models such as linear dynamical systems and stochastic models such Markov decision processes.

For decentralized control the situation is very different. For many problems where the centralized counterpart is simple, such as control to minimize the mean square error, there are no known computationally tractable algorithms for the decentralized case.

We present an overview of what is known, along with our recent results, in which we show that for certain restricted classes of  problems efficient algorithms for finding optimal controller do exist, and for a more general class of systems one may compute approximately optimal controllers efficiently.

Stochastic approximation algorithms are used to approximate solutions to fixed point equations that involve expectations of functions with respect to possibly unknown distributions. Reinforcement learning algorithms such as TD- and Q-learning are two of its most famous applications.

This talk provides an overview of stochastic approximation, with focus on optimizing the rate of convergence. Based on this general theory, the well known slow convergence of Q-learning is explained: the variance of the algorithm is typically infinite.   New algorithms with provably fast (even optimal) convergence rate have been developed in recent years:   stochastic Newton-Raphson,  Zap SNR,  and acceleration techniques inspired by Polyak and Nesterov will be discussed (as time permits, and depending on the interests of the audience).

This work concerns design of control systems for "Demand Dispatch" to obtain ancillary services to the power grid by harnessing inherent flexibility in many loads.  With careful design, the grid operator can harness this flexibility to regulate supply-demand balance.  The deviation in aggregate power consumption can be controlled just as generators provide ancillary service today.  

Distributed control techniques are called for, much like those used today to provide congestion control in communication networks.  The main message is that intelligence should be concentrated as much as possible at the load.  In this way it is possible to design local control loops so that the aggregate of loads appears as a passive input-output system, while strict QoS constraints are maintained for each load.

The inspiration for snake robots comes from biological snakes. Snakes can move over virtually any type of terrain, including narrow and confined locations. They are good climbers, very efficient swimmers, and some snakes can even fly by jumping off
branches and using their bodies to glide through the air. In this plenary talk we will review recent results on modelling, analysis and control of snake robots. The talk will also describe a new research direction within snake robotics, where underwater snake robots are equipped with thrusters along the body to improve maneuverability and provide hovering capabilities, and how this robot addresses current needs for subsea resident robots in the oil and gas industry.

While autonomous agents that perform solo missions can yield significant benefits, greater efficiency and operational capability will be realized from teams of autonomous agents operating in a coordinated fashion. Potential applications for networked multiple autonomous agents include environmental monitoring, search and rescue, space-based interferometers and hazardous material handling. Networked multi-agent systems place high demands on features such as low cost, high adaptivity and scalability, increased flexibility, great robustness, and easy maintenance. To meet these demands, the current trend is to design distributed control algorithms that rely on only local interaction to achieve global group behavior.

The purpose of this talk is to overview our research in distributed control of multi-agent systems. Theoretical results on distributed leaderless consensus with agent dynamics including first- and second-order linear dynamics, rigid body attitude dynamics, and Euler-Lagrange dynamics, distributed single-leader collective tracking with reduced interaction and partial measurements, distributed multi-leader containment control with local interaction, distributed average tracking with multiple time-varying reference signals, and distributed optimization with non-identical constraints will be introduced. Application examples in multi-vehicle cooperative control will also be introduced.

The problem of distributed dynamic state estimation using networked local agents with sensing and communication abilities, has become a popular research area in recent years due to its wide range of applications such as target tracking, region monitoring and area surveillance. Specifically, we consider the scenario where the local agents take local measurements and communicate with only their nearby neighbors to estimate the state of interest in a cooperative and fully distributed manner. A distributed hybrid information fusion algorithm is proposed in the scenario where the process model of the target and the sensing models of the local agents are linear and time varying. The proposed distributed hybrid information fusion algorithm is shown to be fully distributed and hence scalable, to be run in an automated manner and hence adaptive to locally unknown changes in the network, to have agents communicate for only once during each sampling time interval and hence inexpensive in communication, and to be able to track the interested state with uniformly upper bounded estimate error covariance. It is also explored very mild conditions on general directed time-varying graphs and joint network observability/detectability to guarantee the stochastic stability of the proposed algorithm.

Integrated power systems (IPS) incorporate heterogeneous power sources, including energy storage systems, to achieve improved energy efficiency and reliability. They have been a critical enabling technology for vehicle electrification. One distinctive characteristic of IPS is the highly interactive and dynamic nature, due to tight physical couplings of the multiple components involved. To achieve high efficiency, one often exploits their operating profiles and pushes these systems to operate on or close to their admissible boundary, thereby calling for predictive control. In this lecture, we will explore the unique characteristics of the IPS and discuss the challenges and solutions of real-time optimization and predictive control applied to this particular class of systems. Several examples, including the IPS for all-electric ships and the integrated solid oxide fuel cell and gas turbine (SOFC/GT) system, will be used to provide motivations and illustrate the impact of solutions.

Thermal and power systems are tightly coupled and dynamically integrated. The different time scales in thermal and power responses make the integrated thermal and power management problems intriguing and challenging. For connected and automated vehicles (CAVs), the availability of predictive traffic information and the ability to coordinate multiple control subsystems allow us to explore the thermal-power interactions in new dimensions to enhance safety and improve fuel economy. It presents a perfect example where prediction, estimation, control, and optimization serve as the cornerstones for technology breakthroughs in the interconnected and dynamic environment. The talk will discuss the problems, explore the effective tools, and showcase some illustrative solutions.

Nonlinear dynamical systems cover an immensely wide range of real-life situations. However, it is often the case that a priori structure information of the unknown system is not available. Thus, nonparametric identification is necessary for data-driven identification of nonlinear systems. In the first part of this talk, we present a recursive local linear estimator for nonparametric identification of nonlinear autoregressive systems with exogenous inputs. The strong consistency and the asymptotical mean square error properties of the recursive local linear estimator are established, and its application to an additive nonlinear system is discussed. The recursive local linear estimator provides recursive estimates not only for the function values but also their gradients at fixed points. In the second part of this talk, we present a data-driven method for identification of high-dimensional additive nonlinear dynamical systems with little a priori information. In particular, we develop a two-step method for variable selection to determine contributing additive functions and to remove non-contributing ones from the underlying nonlinear system. At the first step, we estimate each additive function by kernel-based nonparametric identification approaches without suffering from the curse of dimensionality. At the second step, we utilize a nonnegative garrote estimator to identify which additive functions are nonzero by use of the obtained nonparametric estimates of each function. We show that the proposed variable selection method can find the correct variables with probability one under weak conditions.

Due to the openness of operation systems and communication interfaces, an increasing number of cyber attacks can easily sneak into Cyber-Physical Systems (CPSs) and cause very serious consequences. Denial-of-service (DoS) attack is one particularly common type of cyber attacks in CPSs. A great deal of efforts have been expended in investigating the effect of power-constrained DoS attacks on the performances of CPSs. Almost all of them assume that communication channel states are unaltered. However, more practically, the variation of communication channel states will impact the consequence of DoS attacks, and thus this factor should not be ignored when investigating the security issues of CPSs. In this talk, we consider the scenario that the sensor data are transmitted through a standard block fading communication channel. From the viewpoint of the DoS attacker, we construct optimization problems considering jointly the system performance indexes and the attack power consumption. Then we transform the original problem into a Markov decision problem and show the existence of optimal solution. As it is difficult to provide an analytical expression of optimal attack strategy, the objective function is approximated to derive an analytical expression of the suboptimal attack strategy.