IEEE.org | IEEE Xplore Digital Library | IEEE Standards | IEEE Spectrum | More Sites
Call for Award Nominations
More Info
In an era where Artificial Intelligence (AI) is often seen as a universal solution for any complex problem, this presentation offers a critical examination of its role in the field of automatic control. To be concrete, I will focus on Optimal Control techniques, navigating through its history and addressing the evolution from its traditional model-based roots to the emerging data-driven methodologies empowered by AI.
The presentation will delve into how the theoretical underpinnings of Optimal Control have been historically aligned with computational capabilities, and how this alignment has shifted over the years. This juxtaposition of theory and computation motivates a deeper investigation into the diminishing relevance of certain traditional control methods amidst the AI revolution. We will critically examine scenarios where AI-driven approaches could outperform classical methods, as well as cases where the hype surrounding AI overshadows its actual utility.
The talk will conclude with a nuanced view of state-of-the-art optimal control methods in practical applications including self-driving cars, advanced robotics and energy efficient systems. From this perspective, we will identify and explore future potential directions for the field, including the design of learning control architectures which seamlessly integrate predictive capabilities at every level, focusing on systems that can autonomously refine their performance over time through continuous learning and interaction with their environment.
The convergence of physical and digital systems in modern engineering applications has inevitably led to closed-loop systems that exhibit both continuous-time and discrete-time dynamics. These closed-loop architectures are modeled as hybrid dynamical systems, prevalent across various technological domains, including robotics, power grids, transportation networks, and manufacturing systems. Unlike traditional “smooth” ordinary differential equations or discrete-time recursions, solutions to hybrid dynamical systems are generally discontinuous, lack uniqueness, and have convergence and stability properties that are defined with respect to complex sets. Therefore, effectively designing and controlling such systems, especially under disturbances and uncertainty, is crucial for the development of autonomous and efficient data-driven engineering systems capable of achieving adaptive and self-optimizing behaviors. In this talk, I will delve into recent advancements in the analysis and design of feedback controllers that can achieve such properties in complex scenarios via the synergistic use of adaptive “seeking” dynamics, robust hybrid control, and decision-making algorithms. These controllers can be systematically designed and analyzed using modern tools from hybrid dynamical systems theory, which facilitate the incorporation of "exploration” and “exploitation" behaviors within complex closed-loop systems via multi-time scale tools and perturbation theory. The proposed methodology leads to a family of provably stable and robust algorithms suitable for solving model-free feedback stabilization and decision-making problems in single-agent and multi-agent systems for which smooth feedback solutions fall short.
Today, it is possible to reprogram the type of a cell for on-demand patient-specific cell therapy, wherein damaged cells in the body are replaced with healthy cells of the correct type generated from easy-to-extract patient’s cells. One approach to produce cells of the desired type is to first reprogram somatic cells, such as skin cells, to pluripotent stem cells, and to then differentiate these pluripotent cells down to the cell type in need. Both processes require accurate control of the temporal concentration of fate-specific proteins, called transcription factors, in the cell in order to efficiently generate high quality output cells. However, so far, accurate control of cellular concentrations has been out of reach. Practitioners inject DNA that produces the appropriate transcription factors in the starting cells at constant rates, without any control on cellular concentrations. In the past decade, the advances in engineering biology have reached the stage where we can implement nonlinear controllers to regulate the cellular level of key molecular players. In this talk, I will illustrate key obstacles to accurate control of protein levels in mammalian cells by conceptualizing the problem through input/output nonlinear, stochastic, models of gene regulation in the context of cell fate determination. I will then use these models to design biomolecular high-gain and integral feedback controllers in mammalian cells to achieve set-point regulation robustly to noise and cellular perturbations. Finally, I will go back to the problem of reprogramming somatic cells to pluripotency and I will show our controllers in action both as a way to uncover optimal reprogramming trajectories and as a way to enforce more accurately optimal transcription factor levels during reprogramming. This is the first instance in which biomolecular controllers have been used for pluripotent stem cell reprogramming. With these tools and experimental demonstrations, we have set the foundations for future research on the use of sophisticated biomolecular networks as controllers of complicated biological processes.
In everyday driving, many traffic maneuvers such as merges, lane changes, passing through an intersection, require negotiation between independent actors/agents. The same is true for mobile robots autonomously operating in a space open to other agents (humans, robots, etc.). Negotiation is an inherently difficult concept to code into a software algorithm. It has been observed in computer simulations that some “decentralized” algorithms produce gridlocks while others never do. It has turned out that gridlocking algorithms create locally stable equilibria in the joint inter-agent space, while, for those that don’t gridlock, equilibria are unstable – hence the title of the talk.
We use Control Barrier Function (CBF) based methods to provide collision avoidance guarantees. The main advantage of CBFs is that they provide easier to solve convex programs even for nonlinear systems and inherently non-convex obstacle avoidance problems. Six different CBF-based control policies were compared for collision avoidance and liveness (fluidity of motion, absence of gridlocks) on a 5-agent, holonomic-robot system. The outcome was then correlated with stability analysis on a simpler, yet representative problem. The results are illustrated by extensive simulations including an intersection example where the (in)stability insights are used to explain otherwise difficult to understand vehicle behaviors.
More information provided here.
Ensemble control deals with the problem of using a common control input to simultaneously steer a large population (in the limit, a continuum) of individual control systems. In this talk, we address a fundamental problem in ensemble control theory, namely, system controllability. A key factor in determining controllability of an ensemble system is its underlying parameterization space. Roughly speaking, the bigger the parameterization space is, the more difficult one can control the ensemble. Over the past two decades, significant progress has been made for understanding controllability of ensemble systems over one-dimensional parameterization spaces, yet little is known when the dimensions are greater than one. A major focus of this talk is to present recent advances in controllability of ensemble systems whose parameterization spaces are multi-dimensional. We will consider two classes of ensemble systems, namely, ensembles of linear control systems and ensembles of control-affine systems. We will first show that linear ensemble systems are problematic if their parameterization spaces are greater than one and, then, show how to resolve this controllability issue by using a special class of control-affine ensembles whose control vector fields are equipped with a fine structure.
Control systems with learning abilities could cost-effectively address societal issues like energy reliability, decarbonization, climate security and enable autonomous scientific discovery. Recent investigations focus on longstanding challenges such as robustness, uncertainty, and safety of complex engineered systems. But most importantly, innovation in deep learning methods, tools, and technology offers an unprecedented opportunity to transform the control engineering practice and bring much excitement to control systems theory research. In this talk, I will introduce recent results in modeling dynamic systems with deep learning representations that embed domain knowledge. I will also discuss differentiable predictive control, a data-driven approach that uses physics-informed deep learning representations to synthesize predictive control policies. I’ll illustrate the concepts with examples from various engineering applications. I’ll close by considering the implications of differentiable programming on the broader control systems context.
Evolution over the course of 500 million years has endowed fish with superior swimming and sensory capabilities in water. This has not only captivated the interest of biologists, but also spurred the development of underwater machines aiming to emulate fish’s locomotion and sensing marvels. In this talk I will first discuss efforts in developing hydrodynamic sensing systems inspired by lateral lines, the flow-sensing organ of fish. I will then illustrate the important role played by advanced modeling and control tools in optimizing robotic fish’s locomotion performance. I will further introduce gliding robotic fish, a new class of robotic fish that incorporates gliding to boost locomotion energy-efficiency, and discuss its application to autonomous underwater sensing. In one example, the unique spiral dynamics of gliding robotic fish is exploited in sampling the distribution of harmful algae along water columns. In another example, a network of gliding robotic fish is proposed for tracking the movement of invasive fish species with acoustic telemetry, and we show how distributed filtering algorithms can be used to localize the moving target. Both examples will be supported with results from field experiments.
The future of healthcare will involve personalized medical therapies for individuals. In applications involving the delivery of a drug (for example, insulin), such personalization can be achieved through the use of tailored feedback control strategies. For close to 30 years, our research group has collaborated with medical experts on the design of algorithms for safe and effective insulin delivery for individuals with Type 1 diabetes mellitus (T1DM). T1DM is a chronic autoimmune disease affecting approximately 35 million individuals world-wide, with associated annual healthcare costs in the US estimated to be approximately $15 billion. Over the years, there have been remarkable innovations in glucose measurement technology, insulin pump design, and personalized control algorithms. Over the last 5 years, multiple commercial closed-loop devices have entered the market, thus delivering the so-called “artificial pancreas” to individuals with T1DM. In this talk, I will outline the difficulties inherent in controlling physiological variables, the challenges with regulatory approval of such devices, and will describe several control systems engineering algorithms we have tested in clinical and outpatient settings for the artificial pancreas. I will describe our work in creating an embedded version of our MPC algorithm to enable a portable implementation in a medical IoT framework and will highlight some of the open challenges for automated insulin delivery. I’ll close by sharing other medical examples where feedback algorithms could provide transformational advances in personalized medicine, including chronotherapy.
People tend to overtrust sophisticated computing devices, especially those powered by AI. As these systems become more fully interactive with humans during the performance of day-to-day activities, ethical considerations in deploying these systems must be more carefully investigated. Bias, for example, has often been encoded in and can manifest itself through AI algorithms, which humans then take guidance from, resulting in the phenomenon of excessive trust. Bias further impacts this potential risk for trust, or overtrust, in that these cyber-physical systems are learning by mimicking our own thinking processes, inheriting our own implicit gender and racial biases, for example. These types of human-AI feedback loops may consequently have a direct impact on the overall quality of the interaction between humans and machines, whether the interaction is in the domains of healthcare, job-placement, or other high-impact life scenarios. In this talk, we will discuss various forms of bias, as embedded in our machines, and possible ways to mitigate its impact on cyber-physical human systems.
Reachability analysis, which considers computing or approximating the set of future states attainable by a dynamical system over a time horizon, is receiving increased attention motivated by new challenges in, e.g., learning-enabled systems, assured and safe autonomy, and formal methods in control systems. Such challenges require new approaches that scale well with system size, accommodate uncertainties, and can be computed efficiently for in-the-loop or frequent computation. In this talk, we present and demonstrate a suite of tools for efficiently over-approximating reachable sets of nonlinear systems based on the theory of mixed monotone dynamical systems. A system is mixed monotone if its vector field or update map is decomposable into an increasing component and a decreasing component. This decomposition allows for constructing an embedding system with twice the states such that a single trajectory of the embedding system provides hyperrectangular over-approximations of reachable sets for the original dynamics. This efficiency can be harnessed, for example, to compute finite abstractions for tractable formal control verification and synthesis or to embed reachable set computations in the control loop for runtime safety assurance. We demonstrate these ideas on several examples, including an application to safe quadrotor flight that combines runtime reachable set computations with control barrier functions implemented on embedded hardware.
The fields of adaptive control and machine learning have evolved in parallel over the past few decades, with a significant overlap in goals, problem statements, and tools. Machine learning as a field has focused on computer-based systems that improve and learn through experience. Oftentimes the process of learning is encapsulated in the form of a parameterized model such as a neural network, whose weights are trained in order to approximate a function. The field of adaptive control, on the other hand, has focused on the process of controlling engineering systems in order to accomplish regulation and tracking of critical variables of interest. Learning is embedded in this process via online estimation of the underlying parameters. In comparison to machine learning, adaptive control often focuses on limited-data problems where fast, on-line performance is critical. Whether in machine learning or adaptive control, this learning occurs through the use of input-output data. In both cases, the approach used for updating the parameters is often based on gradient descent-like and other iterative algorithms. Related tools of analysis, convergence, and robustness in both fields have a tremendous amount of similarity. As the scope of problems in both topics increases, the associated complexity and challenges increase as well. In order to address learning and decision-making in real time, it is essential to understand these similarities and connections to develop new methods, tools, and algorithms.
This talk will examine the similarities and interconnections between adaptive control and optimization methods commonly employed in machine learning. Concepts in stability, performance, and learning, common to both fields will be discussed. Building on the similarities in update laws and common concepts, new intersections and opportunities for improved algorithm analysis will be explored. High-order tuners and time-varying learning rates have been employed in adaptive control leading to very interesting results in dynamic systems with delays. We will explore how these methods can be leveraged to lead to provably correct methods for learning in real-time with guaranteed fast convergence. Examples will be drawn from a range of engineering applications.
Control policies that involve the real-time solution of one or more convex optimization problems include model predictive (or receding horizon) control, approximate dynamic programming, and optimization-based actuator allocation systems. They have been widely used in applications with slower dynamics, such as chemical process control, supply chain systems, and quantitative trading, and are now starting to appear in systems with faster dynamics. In this talk I will describe a number of advances over the last decade or so that make such policies easier to design, tune, and deploy. We describe solution algorithms that are extremely robust, even in some cases division free, and code generation systems that transform a problem description expressed in a high-level domain-specific language into source code for a real-time solver suitable for control. The recent development of systems for automatically differentiating through a convex optimization problem can be used to efficiently tune or design control policies that include embedded convex optimization.
Recent radical evolution in distributed sensing, computation, communication, and actuation has fostered the emergence of cyber-physical network systems. Examples cut across a broad spectrum of engineering and societal fields. Regardless of the specific application, one central goal is to shape the network collective behavior through the design of admissible local decision-making algorithms. This is nontrivial due to various challenges such as the local connectivity, imperfect communication, model and environment uncertainty, and the complex intertwined physics and human interactions. In this talk, I will present our recent progress in formally advancing the systematic design of distributed coordination in network systems. We investigate the fundamental performance limit placed by these various challenges, design fast, efficient, and scalable algorithms to achieve (or approximate) the performance limits, and test and implement the algorithms on real-world applications.
