The Smart Infrastructure Lab is focused on designing real-time control algorithms for operating our societal-scale infrastructure systems through the use of information technology and a more active role for humans in the control loop. Our objective is to improve sustainability, reliability, efficiency, and propose disruptive new methods to address the challenges of rapid urbanization. Some main areas of current focus are as follows:
Energy Management Systems:
To enable a successful integration of renewable technologies and also improve power grid reliability, our lab studies methods to motivate changes in electricity use by end-use customers in response to changes in the availability of supply and the dynamic state of the grid. We design distributed monitoring and control loops to learn customers' behavior and enable them to adjust their electricity demand in real-time, as well as pricing mechan isms for wholesale electricity markets in the presence of flexible demand.
Electric and Driverless Transportation Systems: A growing population of Electric Vehicles (EV), along with disruptive technologies fast happening in the area of driverless cars and shared vehicle fleets, are expected to create synergies that can revolutionize our future transportation systems. In our research, we study the mobility needs of individuals or vehicle fleets in order to design methods that guide EV drivers (or manage autonomous EV fleets) towards using transportation and charging infrastructure systems more efficiently.
Interdependent Networked Infrastructure Systems: Our infrastructure systems depend on each other to function. For example, EV owners introduce couplings between power and transportation systems. Our goal is to provide insights as to how the operations of multiple networked infrastructure systems can be coordinated, and propose solutions that can enhance reliability and efficiency.
Smart cities: Engineers can now access massive amounts of data about the daily activities and needs of citizens as well as the availability of resources in urban environments. Our lab studies dynamic monitoring and resource allocation algorithms that can systematically gather and use this data to better use a city's assets and improve its residents' lives.
Dr. Marden's research focuses on the area of feedback control and systems theory with an emphasis on the development and analysis of game theoretic methods for the coordination of large scale distributed systems. His work is motivated by the fundamental challenges associated with the coordination of multiagent systems with applications including distributed traffic routing, dynamic resource allocation, wind farm control, and sensor networks.
The lab's main areas of research focus on the following three classes of systems:
Autonomous mobile robotic networks can have a tremendous impact in many different areas such as disaster relief, emergency response, and national security. In such networks, a group of unmanned vehicles with limited local sensing, processing, communications, and actuation capabilities are given a task to perform jointly. Each node needs to constantly plan and adapt its trajectory, in order to sense and gather as much information from the environment as possible while maintaining the needed connectivity for cooperation. Its given resources such as energy, time or bandwidth are further very limited, while the operation environment presents uncertainty for sensing, communications and navigation.
Thus, in order to realize the full potentials of a robotic network, we need a foundational understanding of the interplay between sensing, communications and navigation in these systems, which is the main motivation for the ongoing work in our lab. More specifically, we work on developing a new multi-disciplinary paradigm for the optimum design, task assignment and use of limited resources in robotic networks. Our current research thrusts include communication-aware navigation and decision making, compressive sensing and control, obstacle mapping, robotic routers, and cooperative information processing. By utilizing tools from control, robotics, communications, and signal processing, and co-optimizing the sensing, communications and navigation aspects of these problems, we have shown that considerable performance improvement can be achieved.
In one line of work, for instance, we have developed the mathematical foundations of communication-aware motion planning, a term we use to refer to a navigation strategy in which each robot considers not only the impact of motion decisions on its sensing, but also considers realistic probabilistic communication prediction metrics when planning its trajectory. In order for each node to do so, it needs to have an assessment of the link quality at the locations that it has not yet visited. Thus, we have further brought an understanding to the fundamentals of wireless channel predictability, which is key to communication-aware navigation.
In another line of work, we have shown how a robotic network can build an understanding of its environment with minimal sensing and in non-conventional ways, by exploiting the sparse representation of the information in another domain. For instance, we have shown that through proper motion design and by exploiting the sparse representation of the environment, it is indeed possible for the robotic network to see through the walls and build a spatial map of occluded obstacles, using only a small number of wireless measurements.
Overall, our work allows the agents to better understand their environment with minimal sensing, optimize the flow of information between them, and successfully operate under uncertainty and resource constraints.
As computers, digital networks, and embedded systems become ubiquitous and increasingly complex, one needs to understand the coupling between logic-based components and continuous physical systems. This prompted a shift in the standard control paradigm in which dynamical systems were typically described by differential or difference equations to allow the modeling, analysis, and design of systems that combine continuous dynamics with discrete logic. This new paradigm is often called hybrid, impulsive, or switched control.
While some of our work on hybrid systems is of a theoretical nature, it is motivated by several high-impact application areas:
The Center: The Center for Control, Dynamical Systems and Computation (CCDC) was founded in 1991 to facilitate interdepartmental research in control engineering, dynamical systems, and computation and to provide a forum for the exchange of ideas through lectures, seminars, conferences, scholarly meetings, and publications. Currently, the center brings together faculty, students, and visitors from the Departments of Electrical, Mechanical, and Chemical Engineering, Mathematics, Statistics and Applied Probability, and Computer Science.
Activities: The Center focuses on initiating and coordinating research projects with cross-disciplinary investigations and applications to industrial, environmental, transportation, and defense systems. To that end, the Center organizes a weekly seminar series that brings in domain experts, workshops, short courses, publishes a technical report series, and sponsors several fellowships and awards. The Center has a comprehensive graduate program in which students can pursue their studies across several engineering departments. The Center also offers post-doctoral positions and regularly hosts visiting faculty to facilitate inter-institutional and international cooperation.
Cooperation: The Center strives to enable cross-disciplinary teamwork in which theoretical concepts, computational methods, and engineering design are organically merged. Interactions among faculty create an environment in which members benefit from the expertise across several engineering disciplines, applied mathematics and scientific computation. The Center’s collaborative nature is evidenced in the frequent cross listing of courses, as well as joint projects and publications.
Areas of expertise within the center include: nonlinear control, robust control, optimal control, process control, hybrid systems, system identification, stochastic control, distributed control, systems biology, dynamical systems, multi-agent systems, power networks, autonomous vehicles, robotics, and scientific computing.
Research Interests: Robotics, Dynamics, Locomotion, Control, Autonomous Systems, Learning
In our lab group, we study the interplay between passive dynamics and active control in achieving two of the most fundamental challenges in autonomous robotics: locomotion & manipulation.
The three central applications on which we are currently focused are: rough-terrain legged locomotion; maneuverable flapping-wing and rotary flight; and variable-compliance manipulation.
Our overall approach in all work involves developing simplified dynamic models, deriving control laws for achieving particular motions and limit cycle behaviors, and predicting and optimizing reliability. This last aspect is both critical and challenging, because robots in real-world environments are typically subject to significant underactuation and stochasticity. By contrast, the classic study of robotics is founded on having well-characterized and deterministic ("robotic") behaviors from our machines.
Within robotics, our work essentially focuses on the mid-level control problem of getting a desired dynamic response for a body's overall motion through exploitation of both physical impedance (i.e., stiffness, mass and damping) and active control. We invite academic and industry collaboration on both higher-level tasks, such as vision or motion planning, and on low-level problems, such as variable-impedance actuation, real-time control platforms, or novel sensing solutions.
Historically computers were first used for the numerical solution of partial differential equations (PDEs) that arise in engineering design work. Today the desire to create "virtual labs" where engineers can quickly design, test and tune complex designs on a computer, is a major motivator for designing extremely fast, memory efficient algorithms for physical simulations. This is the aim of the Scientific Computing Group.
To design fast algorithms for PDEs we have to exploit special patterns that arise in these problems. The patterns are complicated and their description requires tools from functional analysis, harmonic analysis, approximation theory, elliptic PDE theory, linear algebra and computer science.
We have attacked the problem on several fronts. At the numerical linear algebra end, we have worked on exploiting low-rank micro-structure in matrices to speed-up the solution of large linear systems of equations.
At the functional analysis end, we have worked on developing new breakthrough ideas on high-order linear polynomial approximators, and used these to develop new classes of highly efficient high-order schemes for PDEs.
However, we are still far from delivering a virtual lab-on-a-computer to engineers. There are many examples of PDEs where the current state-of-the-art can be dramatically improved. To achieve this involves the clever synthesis of many existing ideas, and more importantly, new ones that take better advantage of the patterns present. This is an exciting field at the junction of mathematics, computing and physics, and we expect many more breakthroughs ahead.