Daniel E. Rivera - US grants

Process control systems enable plants to meet objectives such as maintaining product quality and minimizing energy consumption, which in turn maximize the plant's profitability and rate-of-return or investment. Because they enable a rapid response to change, process control systems play a vital part in a company's ability to remain profitable in an uncertain economic climate. The first step in the design of an advanced control system is to build a model that represents the dynamics of the plant. Most plants are too complex or the underlying processes too poorly understood to be adequately modeled using first principles. The most reasonable way to obtain reliable dynamic models is from data generated through well designed experiments. In the petrochemical and refining industries, black-box models obtained from experiments are by far the most common means of obtaining dynamic models. The task of obtaining dynamic models from data is referred to as system identification. Control-relevant system identification is motivated by the desire to increase the utility and acceptance of advanced identification concepts in the process industries. It takes into consideration the closed-loop control objectives and the skill level of the user. Control-relevant identification therefore offers the opportunity for the migration of advanced identification concepts to nonexpert users and for the development of computer-aided design tools for identification that can be used by practicing engineers with a B.S. level of education. The main objective of this project is to investigate the subject of control-relevant identification. The basis for the control-relevant approach is the relationship between the design variables of the identification problem and the performance objective of the control problem. To obtain this, bias and variance expressions in the frequency domain and representations of the control problem in terms of linear fractional transformations are used. This analysis leads to a systematic procedure for prefilter design that substantially improves the performance of prediction-error algorithms without demanding substantial increases in skill from the user. In addition, use of the Structured Singular Value leads to a model validation procedure for identified models that provides a clear picture of model limitations to achievable control performance. By providing the theoretical basis for improved computer-aided design tools, these results should make the application of advanced identification concepts a more commonplace practice by engineers in the process industries.

This grant provides funding for the development and evaluation of model predictive control methods for integrating entities in supply chains where business and production conditions vary significantly over time. Increasingly, companies find that they must reconfigure their supply chain structure, policies, and operating conditions to maintain or improve their performance in dynamic, real-time environments. In this research, a Model-on-Demand Predictive Control approach will provide a closed-loop methodology for integrating supply chain decisions that generates desirable system behaviors for the enterprise using both feedback and feedforward control action on the dynamical system. In addition, the model predictive control approach will provide the ability to evaluate the economic benefits of heterogeneous supply structures and varying degrees of centralization and information sharing, and the impact of time-varying inputs and outputs on system performance. Using a series of simulation experiments, the Model-on-Demand Predictive Control approach will be compared with the performance of Materials Requirement Planning logic with a rolling horizon implementation. Measures of performance evaluated in this study include inventory, customer service, responsiveness, and schedule stability.

If successful, the model predictive control approaches developed through this research will provide benefits that include: (1) improved understanding of the value of real-time, feedback controlled information across heterogeneous organizations in supply chains, (2) demonstration of the value of an integrated modeling approach that uses operations research, data warehousing, estimation procedures, and process control in a distributed decision making environment, (3) managerial insights on designing and operating dynamic supply chains for improved performance through increased coordination and information sharing, (4) advancement of the Internet as the enabling technology for closed-loop decision making across organizational boundaries, and (5) foundation for future research on supply chain integration using dynamic control approaches in real-time environments such as those found on the Internet including supply web enterprises and business-to-business marketplaces.

A vibrant manufacturing sector is vital to the economic health of the United States, and efficient management of supply chains plays a critical role in this regard. This grant provides funding to explore how Model Predictive Control (MPC), an advanced control paradigm originating from the process industries, can be utilized as a novel approach for tactical decision-making in supply chain management problems associated with semiconductor manufacturing. The project involves the unique collaboration of investigators from chemical engineering, mathematics, and computer science at Arizona State University with a technical leader from Intel Corporation. The principal topic of research is the formulation and initial proof-of-concept of Model Predictive Control algorithms for a novel class of supply chain problems associated with semiconductor manufacturing that extend beyond chemical process applications and traditional MPC. To this end, the use of state-of-the-art optimization techniques that efficiently solve the optimization problems associated with Model Predictive Control, and the development of a software architecture that can satisfy the unique computational needs of this class of supply chain problems, will be examined.

It is expected that the Model Predictive Control-based formulations developed as part of this research will ultimately serve as integral components in hierarchical, enterprise-wide planning tools that function on real-time data, support varying levels of information sharing and centralization, and employ combined feedback-feedforward control action. Broader impacts of this research include efforts towards developing a body of theory and technology meaningful not only to semiconductor manufacturing, but to a wide range of discrete-part manufacturing industries of importance to the national economy. Ongoing efforts by the PIs to involve minority and undergraduate students in integrative research and educational activities, as well as the dissemination of research outcomes to the engineering community at large, will be further stimulated as a result of this grant.

[unreadable] DESCRIPTION (provided by applicant): I am a chemical engineer whose research career has spanned the study of control engineering concepts in diverse application settings, from chemical process control to supply chain management to (more recently), adaptive interventions for the prevention and treatment of drug abuse. Adaptive interventions systematically individualize therapy through the use of decision rules that act on measurements of tailoring variables over time. I seek a K25 Mentored Quantitative Research Career Development award for the purpose of establishing myself as an independent researcher in the field. Control systems are used in engineering applications as a means to transform the behavior of a system over time from undesirable conditions to desirable ones; my work to date has established that adaptive interventions represent a form of feedback control in the context of behavioral health. Consequently, drawing from ideas in control engineering has the potential to significantly inform the analysis, design, and implementation of these interventions, leading to improved adherence, better management of limited resources, a reduction of negative effects, and overall more effective interventions. My research activities as part of this award, under the mentorship of Linda Collins (Penn State) and Susan Murphy (Michigan), and in collaboration with scientists affiliated with the Prevention Research Center at Arizona State (led by Irwin Sandier) and the Center for Continuum of Care in the Addictions at Penn (led by James McKay), will expand upon conceptual connections between adaptive interventions and control engineering principles by developing realistic simulation testbeds involving the prevention and treatment of multiple co-occurring disorders associated with substance use, HIV/AIDS, and mental health. The simulations will be used to better understand how to effectively integrate decision rules in a clinical context, and will serve as a basis to extend to problems in drug abuse prevention and treatment two significant engineering disciplines that form an important part of my expertise: modeling of phenomena associated with drug abuse using system identification methods, and optimized decision policies for multi- component interventions based on the concept of Model Predictive Control. The opportunity afforded by this award for significant interaction with prevention scientists and leaders in the field will insure that the outcomes of this research remain grounded in reality and have practical significance. [unreadable] [unreadable] [unreadable]

DESCRIPTION (provided by applicant): This study will develop methods to enhance the conduct of research in the area of behavioral intervention development and evaluation. Behavioral interventions aim to prevent and treat disease by using a strategy that relies on reducing unhealthful behaviors and promoting healthful behaviors. These interventions play an increasingly prominent role in a wide variety of areas of public health importance, including drug abuse, HIV/AIDS, cancer, mental health, diabetes, obesity, cardiovascular health, and aging. The standard treatment/control randomized clinical trial (RCT) provides a principled methodological framework for establishing whether behavioral interventions work. The proposed research will develop a corresponding principled methodological framework for building interventions that have been optimized so that they are operating at peak efficacy (impact under ideal conditions), effectiveness (impact in real-world conditions) and efficiency (impact in relation to use of resources). The interdisciplinary research team includes a behavioral scientist and an engineer as PI's, statisticians, and a distinguished panel of eight behavioral intervention scientists from different public health areas. The proposed framework for optimizing behavioral interventions is based on methods widely used in engineering. This research will adapt these methods for use in behavioral interventions. The methods involve expressing behavioral interventions as detailed dynamical models. Dynamical models are well suited to behavioral interventions, which are typically complex multivariate multi-level time-varying processes. After a dynamical model of a behavioral intervention has been expressed, it can then used as part of established engineering procedures to optimize the intervention. This project has three Specific Aims. The first is to work with each member of the panel of behavioral intervention scientists to express an intervention as a detailed dynamical system model, and then to apply engineering optimization methods, such as Internal Model Control and Model Predictive Control, to each one. The second Specific Aim is to develop, document, and disseminate a computer program that behavioral scientists can use to model behavioral interventions as dynamical systems and apply optimization techniques to them. The third Specific Aim is to lay the groundwork for further adaptation of engineering optimization approaches for use in behavioral science. This part of the project will focus on system identification and multi-level optimization. Benefits of the proposed research will extend to any area of public health that employs behavioral interventions for prevention and treatment of disease, because it will result in behavioral interventions that are more efficacious, effective, and efficient at reducing morbidity and mortality. The proposed work will lead directly to improved behavioral interventions for prevention and treatment of disease. Any area of public health that employs behavioral interventions will benefit from the resulting increase in intervention efficacy, effectiveness and efficiency and corresponding reduction in morbidity and mortality.

EAGER: Defining a Dynamical Behavioral Model to Support a Just in Time Adaptive Intervention

This EAGER supports exploratory work to develop a novel approach to the creation of a dynamical computational model of human health behavior. The goal of this high risk project is to apply an experimental design that merges methods from multiple disciplines to generate the necessary data to develop a dynamical systems model of human health behavior. In theory, mobile technologies have this capacity to provide health interventions in real-time that are adapted to the individual, but in practice the specific theoretical models and decision rules required to determine exactly when, where, and how to intervene do not exist. Standard health approaches use theoretical frameworks to identify and select target behaviors and approaches or intervention. By creating dynamical models of human behavior, real-time adaptive interventions can be developed and empirically assessed building the foundational science of computational behavior. While this project is concerned with creating dynamic computational models for increasing exercise behavior, the approach may find applications more broadly with a wide range of human health issues.

The goal of this EAGER project is to create a mathematical model that will provide the evdience for making decisions about when, where, and how a "just in time" adaptive mHealth physical activity intervention should be implemented. Creating this dynamical behavioral model is a challenging problem that requires insights from different disciplines because behavioral science provides insights regarding what to measure, and behavioral intervention strategies that could be used dynamically; however, current behavioral theories fail to provide any real insights on when, where, and how to intervene at the opportune moment. Control systems engineering provides a methodology for creating dynamic mathematical models and decision-making, but this methodology has only sparsely been applied in a human behavioral context. A key first step for developing a dynamical behavioral model is to gather "informative" empirical data to estimate the model. These data will be generated with an informative system identification "informative" experiment within a human context that builds on lessons from behavioral science about experimental designs and that takes full advantage of the temporally rich data available from mHealth technologies. We will use these data to develop a fundamental yet empirically- supported dynamical behavioral model for understanding our target behavior.

Project Summary Our prosed study will address critical gaps in the literature and practice of informed consent in digital health research. We will leverage the existing Digital Health Checklist (DHC) tool by expanding the consent prototype building component to incorporate what is meaningful to research participants. This study involves co-designing a meaningful informed consent prototype with participants to produce and test a digital health consent blueprint to increase capacity for understanding the function of algorithms used in behavioral interventions. These advances in the DHC tool will contribute to the evidence-base to support the process of informing prospective participants about digital health research. This study will leverage an established decision support tool developed for digital health researchers. The DHC was informed through an iterative design process involving behavioral scientists, regulators, IRB members, ethicists, and clinician-researchers and is grounded in accepted principles of research ethics, namely respect for persons, beneficence and justice, and incorporates four orthogonal domains including: (1) Access and Usability, (2) Risks and Benefits, (3) Privacy, and (4) Data Management. Inspired by an effectiveness-implementation design process, we will test and co-design an interactive consent form with prospective research participants. This human centered participatory design approach will expose unique concerns when asked to use a digital technology to gather personal health information. The proposed work will systematically study and actively respond to critical ethical, legal/regulatory and social implications (ELSI) applied to digital health research - specifically our ability to convey accessible study information such that informed consent transpires. This research will directly benefit our parent R01, will contribute to the literature on informed consent and have potential implications for other personalization algorithms for behavior change, such as those used in industry. Co-designing innovative decision support tools that can be used by researchers, algorithm developers, IRBs, and participants will foster shared decision making at the earliest stages of digital health research and algorithm creation.

A wide range of technologies, such as smartphones, wearables (e.g., Fitbit, Apple Watch), and medical devices use alerts to inspire actions of users. Potentially useful alerts come at the cost of alert fatigue whereby individuals ignore alerts over time. For example, several physical activity interventions use alerts to inspire activity; notifications work initially but with diminished efficacy over time. Ignoring alerts is problematic in a variety of domains. For example, notification fatigue reduces the potency of interventions (e.g., notifications to inspire walking) and can be a safety risk in other areas such as in hospitals where notification fatigue can lead providers to ignore safety alerts (e.g., cross-drug interaction) provided by the electronic medical record. There is a need for novel solutions for reducing alert fatigue. Location, digital traces, and other data enable inference of states when a person would desire/need alerts, henceforth labeled just-in-time states, but more advanced analytics are needed. For example, a suggestion to walk (e.g., SMS saying, Want to go for a walk?) may only produce the desired outcome when a person's state (e.g., low stress) and context (e.g., no meetings, nice weather) align such that the person appreciates the notification (what we label receptivity) and can act on it (what we label opportunity). Estimating the likelihood that a given moment is a just-in-time state requires not only data but also an approach to manage the multivariate, dynamic, idiosyncratic, and multi-timescale nature of the problem. Returning to the walking example, stress, weather, and location change dynamically with each influencing the likelihood that a notification will inspire walking. In our work, results suggest idiosyncrasy in the factors that predict steps: some people walk more when stressed, others less, and still others are not influenced by stress. Further, just-in-time notifications cannot be viewed in a vacuum and, instead, are often part of a more long-term process, such as sustained engagement in a health behavior, thus making it a multi-timescale problem. The purpose of this work is to develop a just-in-time state estimation strategy and to stage a multi-timescale controller for walking as a concrete use-case of a control systems approach to counteract alert fatigue.