This project explores balancing performance considerations and power consumption in cyber-physical systems, through algorithms that switch among different modes of operation (e.g., low-power/high-power, on/off, or mobile/static) in response to environmental conditions. The main theoretical contribution is a computational, hybrid optimal control framework that is connected to a number of relevant target applications where physical modeling, control design, and software architectures all constitute important components. The fundamental research in this program advances state-of-the-art along four different dimensions, namely (1) real-time, hybrid optimal control algorithms for power management, (2) power-management in mobile sensor networks, (3) distributed power-aware architectures for infrastructure management, and (4) power-management in embedded multi-core processors. The expected outcome, which is to enable low-power devices to be deployed in a more effective manner, has implications on a number of application domains, including distributed sensor and communication networks, and intelligent and efficient buildings. The team represents both a research university (Georgia Institute of Technology) and an undergraduate teaching university (York College of Pennsylvania) in order to ensure that the educational components are far-reaching and cut across traditional educational boundaries. The project involves novel, inductive-based learning modules, where graduate students team with undergraduate researchers.

Large-scale applications of cyber-physical systems (CPS) such as commercial buildings with Building Automation System (BAS)-based demand response (DR) can play a key role in alleviating demand peaks and associated grid stress, increased electricity unit cost, and carbon emissions. However, benefits of BAS alone are often limited because their demand peak reduction cannot be maintained long enough without unduly affecting occupant comfort. This project seeks to develop control algorithms to closely integrate battery storage-based DR with existing BAS capabilities. The overarching objective is to expand the building's DR capabilities, providing crucial benefits towards smarter grids, while maintaining appropriate occupant comfort and reducing building ownership cost. This project follows a 2-phase approach towards more effective integration. First, building peak demand forecasting will be added to existing battery dispatch methods. Under electricity tariffs geared towards daily [monthly] peaks, such forecasting could result in the same battery-enabled demand charge (dollars per kW) savings as previously demonstrated storage dispatch algorithms. However, supply charges (dollars per kWh) and associated emissions would be reduced because battery dispatch would be geared towards reducing only the biggest daily [monthly] peaks while not incurring roundtrip charging losses on more moderate peaks. Phase 2 builds on phase 1, adding closer integration and systematic optimization to the algorithms for forecasting, BAS, and battery dispatch. This integration will allow the integrated CPS to manipulate the BAS process itself, thereby optimizing, e.g., light dimming, temperature set-points, and pre-cooling in unison with battery-based DR. Feasibility and future promise of such experimental control methodology will be measured by a multi-objective cost function which includes demand and energy charges, savings from DR participation, storage equipment capital expenditure (required size, achievable lifetime), and occupant comfort. Integrating BAS- with battery-based DR is nascent, mostly because the peak demand forecast, BAS, and storage dispatch algorithms that such a CPS requires have yet to be developed. This project seeks to lay important methodological groundwork for such applications, thus furthering commercial buildings' role in the Internet of Things. The PI's participation in the NIST/US-Ignite Global City Team Challenge (with partners Urban Electric Power, Siemens Corporate Technology, City University of New York, and NY-Best) furthers public engagement with such technology and will help catalyze its translation into the commercial space.

Harnessing wind energy is one of the pressing challenges of our time. The scale, complexity, and robustness of wind power systems present compelling cyber-physical system design issues. Leveraging the physical infrastructure at Purdue, this project aims to develop comprehensive computational infrastructure for distributed real-time control. In contrast to traditional efforts that focus on programming-in-the-small, this project emphasizes programmability, robustness, longevity, and assurance of integrated wind farms. The design of the proposed computational infrastructure is motivated by, and validated on, complex cyber-physical interactions underlying Wind Power Engineering. There are currently no high-level tools for expressing coordinated behavior of wind farms. Using the proposed cyber-physical system, the project aims to validate the thesis that integrated control techniques can significantly improve performance, reduce downtime, improve predictability of maintenance, and enhance safety in operational environments. The project has significant broader impact. Wind energy in the US is the fastest growing source of clean, renewable domestically produced energy. Improvements in productivity and longevity of this clean energy source, even by a few percentage points will have significant impact on the overall energy landscape and decision-making. Mitigating failures and enhancing safety will go a long way towards shaping popular perceptions of wind farms -- accelerating broader acceptance within local communities. Given the relative infancy of "smart" wind farms, the potential of the project cannot be overstated.

It is now recognized that cloud data centers are a significant consumer of energy resources and a substantial source of greenhouse gas emission. On the other hand, the intermittency and uncertainty of renewable energy present a daunting operating challenge for the electricity grid. The key idea behind this CRII: Cyber-Physical Systems project is that these two challenges are in fact symbiotic: data centers can be virtual batteries for the electricity grid. Specifically, data centers are large loads, but are also flexible. If the electricity grid can call on the flexibilities of data centers via appropriate demand response programs, this will be a crucial tool for easing the incorporation of renewable energy into the electricity grid. Unfortunately, despite the great potential, the current reality is that data centers perform little demand response. This project aims at the interdisciplinary challenges of enabling demand response from cloud data centers to realize the societal benefits. The overarching goal of this project is to develop an intellectual framework to understand and guide the realization of demand response from cloud data centers, to address engineering and economic challenges in order to manage the daunting risk. This project will first quantify the potential economic and environmental benefits of demand response from cloud data centers. The quantification includes the societal cost savings and emission reductions from networked data centers through geographical load balancing, and the impacts of demand response taxonomy. Built upon the first thrust, this project will continue to tackle the interdisciplinary challenges of both local control algorithm design and global market design for data center demand response in order to facilitate their participation in various demand response programs. The researchers will study prediction-based pricing design and analysis, demand response program design based on optimization decomposition, and distributed online algorithm design for risk management and distributed control. The results of this project will, at the societal level, help utility companies and load serving entities realize the great potential that lies in the Cloud, and, furthermore, design demand response programs that provide right incentives for data center operators to participate. At a local level, this project will help guide the management of geographically distributed data centers in participating in the right demand response programs. The control algorithms and demand response programs, as well as the methodology, can be applied beyond data centers. This research will create new knowledge in distributed online algorithm design and optimization-based market design. Additionally, this project will help design an interdisciplinary course Sustainable IT and IT for Sustainability. Personnel involved in this project, graduates and undergraduates, will receive innovation experiences through the algorithm design, analysis, implementation, and testing.

This proposal will establish a framework for developing distributed Cyber-Physical Systems operating in a Networked Control Systems (NCS) environment. Specific attention is focused on an application where the computational, and communication challenges are unique due to the sheer size of the physical system, and communications between system elements include potential for significant losses and delays. An example of this is the power grid which includes large-scale deployment of distributed and networked Phasor Measurement Units (PMUs) and wind energy resources. Although, much has been done to model and analyze the impact of data dropouts and delay in NCS at a theoretical level, their impact on the behavior of cyber physical systems has received little attention. As a result much of the past research done on the `smart grid' has oversimplified the `physical' portion of the model, thereby overlooking key computational challenges lying at the heart of the dimensionality of the model and the heterogeneity in the dynamics of the grid. A clear gap has remained in understanding the implications of uncertainties in NCS (e.g. bandwidth limitations, packet dropout, packet disorientation, latency, signal loss, etc.) cross-coupled with the uncertainties in a large power grid with wind farms (e.g. variability in wind power, fault and nonlinearity, change in topology etc.) on the reliable operation of the grid. To address these challenges, this project will, for the first time, develop a modeling framework for discovering hitherto unknown interactions through co-simulation of NCS, distributed computing, and a large power grid included distributed wind generation resources. Most importantly, it addresses challenges in distributed computation through frequency domain abstractions and proposes two novel techniques in grid stabilization during packet dropout. The broader impact lies in providing deeper understanding of the impact of delays and dropouts in the Smart Grid. This will enable a better utilization of energy transmission assets and improve integration of renewable energy sources. The project will facilitate participation of women in STEM disciplines, and will include outreach with local Native American tribal community colleges This project will develop fundamental understanding of impact of network delays and drops using an approach that is applicable to a variety of CPS. It will enable transformative Wide-Areas Measurement Systems research for the smart grid through modeling adequacy studies of a representative sub-transient model of the grid along with the representation of packet drop in the communication network by a Gilbert model. Most importantly, fundamental concepts of frequency domain abstraction including balanced truncation and optimal Hankel-norm approximation are proposed to significantly reduce the burden of distributed computing. Finally, a novel `reduced copy' approach and a `modified Kalman filtering' approach are proposed to address the problem of grid stabilization using wind farm controls when packet drop is encountered.

This CAREER project responds to an urgent need to develop mobile power distribution systems that lower deployment and operating costs while simultaneously increasing network efficiency and response in dynamic and often dangerous physical conditions. The significant need for an efficient and effective mobile power distribution system became evident during search and rescue/recovery missions following the Japan tsunami and the disappearance of the Malaysia MH370 airplane. The technology outcomes from this project will apply to a broad range of environments (in space, air, water or on ground) where the success of long-term robotic network missions is measured by the ability of the robots to operate, for an extended period of time, in highly dynamic and potentially hazardous environments. These advanced features will provide the following advantages: efficiency, efficacy, guaranteed persistence, enhanced performance, and increased success in search/rescue/recovery/discovery missions. Specifically, this project addresses the following technology problems as it translates from research discovery toward commercial application: inflated energy use currently required when the autonomous vehicles break from mission to return to recharging station; lack of multi-robot coordination needed to take into account both fundamental hardware and network science challenges necessary to respond to energy needs and dynamic environment conditions. By addressing these gaps in technology, this work establishes the theoretical, computational, and experimental foundation for mobile power delivery and onsite recharging capability. Moreover, the new technology developed in this project is universally adaptable for disparate autonomous vehicles especially autonomous underwater vehicles (AUVs). In more technical terms, this project creates network optimization and formation strategies that will enable a power distribution system to reconfigure itself depending on the number of operational autonomous vehicles and recharging specifications to meet overall mission specifications, the energy consumption needs of the network, situational conditions, and environmental variables. Such a system will play a vital role in real-time controlled applications across multiple disciplines such as sensor networks, robotics, and transportation systems where limited power resources and unknown environmental dynamics pose major constraints. In addition to addressing technology gaps, undergraduate and graduate students will be involved in this research and will receive interdisciplinary education/ innovation/ technology translation/ outreach experiences through: developing efficient network energy routing, path planning and coordination strategies; designing and creating experimental test-beds and educational platforms; and engaging K-12th grade students in Science, Technology, Engineering and Math including those from underrepresented groups. This project engages Michigan Tech's Great Lake Research Center (GLRC) and Center for Agile Interconnected Microgrids (AIM) to develop experimental test-beds and conduct tests that validate the resulting methods and algorithms, and ultimately, facilitate the technology translation effort from research discovery toward commercial reality.

The electric power grid, a cyber-physical system (CPS), faces an alarmingly high risk of catastrophic damage from cyber-attacks. However, modeling cyber-attacks, evaluating consequences, and developing appropriate countermeasures require a detailed, realistic, and tractable model of electric power CPS operations. The primary barrier is the lack of access to models for the complex legacy proprietary systems upon which the electric power grid has relied for decades. This project aims to overcome these challenges with the development of an attack-verifying (verifiable) software framework that will capture the electric power system operations in adequate detail. Cyber threats will be verified using this framework through a combination of sound theoretical methods and an open-source commercial simulation engine accessible via a unique transition to practice (TTP) option. This research focuses on four fundamental and related thrusts: (i) identifying classes of cyber-attacks with quantifiable physical consequences and developing detection-based countermeasures; (ii) identifying communication attacks on distributed grid operations and developing information-sharing countermeasures; (iii) developing a verifiable software framework that models the spatio-temporal operations of the electric grid in tandem with thrusts (i) and (ii) to verify attack models, evaluate countermeasures, and develop new resiliency protocols; and (iv) a TTP option, in collaboration with industry-leading experts from IncSys and PowerData, to develop commercial grade open source power simulation software packages to integrate and test the attacks and countermeasures of Thrusts (i) through (iii) as well as develop workforce training curriculum for North American Electric Reliability Council (NERC) certification. This research also includes engagement with K-12 students via the Arizona Science Laboratory program.

The electric power grid is a complex cyber-physical system (CPS) that forms the lifeline of modern society. Cybersecurity and resiliency of the power grid is of paramount importance to national security and economic well-being. CPS security testbeds are enabling technologies that provide realistic experimental platforms for the evaluation and validation of security technologies within controlled environments, and they also enable the exploration of robust security solutions. The project has two objectives: (a) to develop innovative architectures, abstractions, models, and algorithms for large-scale CPS security testbeds; and (b) to design and implement a high-fidelity, scalable, open-access CPS security testbed for the smart grid, and to conduct research experimentation. The testbed integrates appropriate cyber-control-physical hardware/software components, models, and algorithms in a modular design that enables federation of smaller testbeds to form a large-scale virtual experimental environment. The use cases for the testbed include vulnerability assessment, risk assessment, risk mitigation studies, and attack-defense exercises. The project also aims to develop standardized datasets, models, libraries, and use cases, and make the testbed available to a broader research community through an open-, remote-access model by leveraging collaboration from academic and industry partners. Besides contributing to research and technology that will enable a future electric power grid that is secure and resilient, this project develops and disseminates innovative curriculum modules including CPS Cyber Defense Competitions (CPS-CDC) for imparting security knowledge to students via an inquiry-based learning paradigm. The project also mentors students, including underrepresented minorities, in thesis work and Capstone projects, and exposes high-school students to cybersecurity concepts via testbed demonstrations.

One of the challenges for the future cyber-physical systems is the exploration of large design spaces. Evolutionary algorithms (EAs), which embody a simplified computational model of the mutation and selection mechanisms of natural evolution, are known to be effective for design optimization. However, the traditional formulations are limited to choosing values for a predetermined set of parameters within a given fixed architecture. This project explores techniques, based on the idea of hidden genes, which enable EAs to select a variable number of components, thereby expanding the explored design space to include selection of a system's architecture. Hidden genetic optimization algorithms have a broad range of potential applications in cyber-physical systems, including automated construction systems, transportation systems, micro-grid systems, and space systems. The project integrates education with research by involving students ranging from high school through graduate school in activities commensurate with their skills, and promotes dissemination of the research results through open source distribution of algorithm implementation code and participation in the worldwide Global Trajectory Optimization Competition. Instead of using a single layer of coding to represent the variables of the system in current EAs, this project investigates adding a second layer of coding to enable hiding some of the variables, as needed, during the search for the optimal system's architecture. This genetic hiding concept is found in nature and provides a natural way of handling system architectures covering a range of different sizes in the design space. In addition, the standard mutation and selection operations in EAs will be replaced by new operations that are intended to extract the full potential of the hidden gene model. Specific applications include space mission design, microgrid optimization, and traffic network signal coordinated planning.