Selected Grants:

Goals and Results Described Below the List

  1. National Institutes of Health – NEI R01, “Personalized Forecasting of Disease Trajectory for Patients with Open Angle Glaucoma,” $2,883,323. 9/1/2016 to 8/31/2021. Co-PIs:  Joshua Stein of Ophthalmology and Mariel Lavieri of Industrial and Operations Engineering. I was a co-author of proposal with the other two authors:  Stein and Lavieri.  Co-Investigators: Mark Van Oyen of Industrial and Operations Engineering, Christopher Andrews of Ophthalmology, Mae Gordon (Professor in the Department of Ophthalmology and Visual Sciences, Washington University of St. Louis), Chris Johnson (Professor in the Department of Ophthalmology and Visual Sciences, University of Iowa), Joel Schuman (Chair of the Department of Ophthalmology at New York University), Gadi Wollstein (Professor of Ophthalmology and Bioengineering, University of Pittsburgh). My share (direct+indirects): $539,109.
    1. Abstract – click here. 
    2. The goal of this study is to refine and enhance a forecasting tool that we have begun developing to assist eye doctors (a) by helping to identify which patients will develop glaucoma or will experience worsening of existing glaucoma, and at what pace, (b) by recommending when the patient should next be assessed for possible disease worsening, and (c) by calculating the patient’s optimal intraocular pressure, a vital measurement in glaucoma that can be controlled through carefully targeted treatment. These highly accurate forecasts and recommendations, which would be extremely difficult for an eye doctor to calculate correctly using his brain power alone, are calculated using sophisticated engineering and mathematical techniques, incorporating valuable detailed information from previous and ongoing government-sponsored treatment trials and new and previously obtained data on the specific patient for whom the forecasts and recommendations are being made. This tool will greatly inform doctors’ decisions on who, when, and how aggressively to treat, in a way that avoids overtreatment and unnecessary treatment but gives the patients at highest risk for blindness their best possible chance at preserving their long-term sight, thus fulfilling a critical mission of the National Eye Institute.
  2. MTRAC for Life Sciences Kickstart Award UM IR #5140 “Monitoring Patients with Glaucoma Using a Novel Personalized Forecasting Tool” $26,000 4/25/2016 to 4/25/2017 co-Investigator (J.D. Stein PI) , My Share 33%.
    • Developed and implemented a dynamic decision support tool for monitoring glaucoma that was featured in EyeNet (Am. Acad. Ophth.) as one of the top 4 innovations in glaucoma (2014).  As V1.0, with a large team of students and professionals (supported by a UM translational grant), we developed a machine-learning algorithms to predict disease progression risk and optimize the timing of subsequent testing. These dynamic methods may improve how rapidly disease progression is identified by 40-60%. We helped Apervita Corp. to host our algorithms in their cloud platform to enable widespread adoption. A web-based micro-site interfaced with Apervita to reach patients with glaucoma.
  3. National Science Foundation, “EAGER: Advanced Capacity Allocation Methodology: Time-sensitive Appointments in Congested Service Systems,” $242,072. 1, 2015–Aug. 31, 2017, PI: M.P. Van Oyen. My share: 100%. Supports 1 Ph.D. student year-round and 1 undergraduate in summer.
    • The primary goal of the research was to generate new methodologies to achieve better health outcomes for healthcare delivery systems and better patient experiences. Because patient health outcomes are adversely affected by long waits from the time of request or referral for specialist care until the patient’s appointment day, particular attention was given to methods to increase the efficiency and effectiveness of methods for planning and scheduling.

      This research developed novel methods to improve the ability to provide more rapid access to urgent patients so urgent patients don’t wait so many days for a care appointment as less urgent patients do. In addition, our advances to the state of the art of optimization based scheduling and planning also allows reduced access delays while maintaining or improving the utilization of care providers. In integrated healthcare systems, patients with complex conditions require multiple visits in various departments over time as one of many complicating aspect that this research considers. Current practice in healthcare delivery tends to use a basic first come first served rule when setting appointment dates. While first come first served scheduling is simple and familiar, it fails to enable appointment setting that gives relatively lesser waits to the relatively more urgent patients. This research addressed this gap by creating methods that allow providers to service multiple classes of patient types and offer each type of patient the expectation that their waiting times will only rarely exceed a promised level. The proposed methods will set appointments by date so that the lengths of delays that are consistent with the urgency needs of that class of patients. This is mathematically difficult, because these various type of patients share key service resources, such as physicians. Resource utilization, staff overtime, volume of services fulfilled, and waits to receive an appointment are all considered in the modeling methods developed. It is common practice for physicians to have a planned work template that describes the allocation of their time to types of care visits. The methodology developed can optimize such templates (at an individual or departmental level) to achieve these daily appointment scheduling changes.

      The research was conducted in collaboration with several health care systems to demonstrate that significant benefits can be achieved in several important areas of healthcare delivery. Some of the first ever system optimization models were invented to achieve timely delays to receiving care for various patient type being serviced by a network of outpatient clinics. Very little rigorous operational research has incorporated the issue of coordinating care visits so that care visits occur in the correct order with appropriate clinicians and in a timely manner as well. Our research included the challenge of the scheduling of a clinic visits to be followed by a surgery within a time window dictated by the type of surgery and severity of the case. When only a modest delay for surgery is allowed for the more urgent types, a tremendous amount of overtime is required with traditional scheduling. In cases studies, the overtime was reduced by 30-50% while achieving the access delay maximum wait times, which can vary from 3 days to 40 days. One of the approaches to urgency based scheduling that was published thanks in part to the funding of this grant was subsequently modified and implemented by a prominent healthcare provider with positive benefits.

      The improved operational decision support methods developed meet the need of healthcare provision organizations to (1) reduce costly overtime, (2) improve access to care by better utilizing care providers, (3) reduced wait to get an appointment, and (4) facilitate coordinated clinical care visits.

      Case studies of multiple healthcare delivery setting were performed and published or submitted for publication to illustrate how cutting edge research can address these major needs in appointment scheduling and resource allocation.  The results of this work have been incorporated into university courses that include Practicum in Hospital Systems, Simulation Analysis and Design, and Queueing Networks. The research benefited from the unbudgeted collaborative efforts of multiple faculty, physicians, and medical staff at multiple health institutions including the University of Michigan, St. Joseph Mercy Hospital, Memorial Sloan Kettering Cancer Center, and Mayo Clinic.

  4. Glaucoma Research Foundation (GRF) Shaffer Grant, “A Dynamic, Personalized Glaucoma Monitoring Decision Support Tool,” $40,000. Feb 15, 2014– Feb 14, 2015, PI: Joshua D. Stein (co-I’s M.P. Van Oyen and M.S. Lavieri) Entire grant devoted to supporting Pooyan Kazemian, Ph.D. student for 2.5 Terms (10 months) March – December 2014.  
    •  The primary focus of this study is open-angle glaucoma, a major cause of blindness worldwide.  This grant played a role in helping us to imporove glaucoma decision support using a type of Machine Learning/AI.  Our work advanced the theory of “measurement subsystems control” and used it to recommend the degree of intervention as well as the costly measurements to take so as to optimize the tradeoffs over a future horizon through LQG control.
  5. National Science Foundation, “Stochastic Modeling and Optimization of Longitudinal Health Care Coordination,” $420,000. 1, 2012–Aug. 31, 2016, PI: M.P. Van Oyen. My share: $367,350. Supports 1 Ph.D. student year-round and 1 undergraduate in summer.
    • The primary goal of the research was to generate new methodologies for improving operational planning and scheduling in healthcare environments, with applications for service sector operations more generally. Particular attention was given to methods to increase the efficiency and effectiveness of (1) clinical research units which provided shared infrastructure for the delivery of clinical trials for newly developed drugs and treatments, (2) operational systems that facilitate the integration of clinical research with clinical care, (3) healthcare that is longitudinal in nature, requiring coordinated services over time.

      Clinical research units (CRUs) are extremely complex and important, and particular focus is given to advanced decision support models and methods to enable better appointment scheduling, allocation of staff/nurses to patient visits, and other decisions surrounding capacity and which new clinical trials can be conducted with available resources without undermining the clinical trials already under way. This research is the first to address the issues of trial selection, the stochastic enrollment of participants, appointment date setting, and the joint optimal allocation of rooms/equipment and nurses in CRUs.

      The work developed methods of decision support for optimizing the mix of clinical trials that maximizes the value of a clinical research unit while maintaining good service levels to the trials conducted. This research is the first to tackle the operational issues of how to best manage clinical research units and other departments while including key features such as staff size, number of rooms, cross-training strategies, decisions on which potential clinical trials or services can be offered, and methods for setting appointment dates. It also provides methods for effectively integrating clinical research activities into the portfolio of clinical care provided by a healthcare service.

      In the direction of longitudinal care, our research successfully developed complex optimization models that improve the access to and utilization of critical healthcare resources such as intensive care units (ICUs) and “stepdown” or Intermediate Care Units (IMCs). It better supports evidence-based medicine by systematizing the admission process for patients using a novel mortality risk metric unique to the patient’s history. The decision of whether to admit a patient to a critical care unit is a crucial operational problem that has significant influence on both hospital performance and patient outcomes. Hospitals currently lack a methodology to selectively admit patients to these units in a way that patient health risk metrics can be incorporated while considering the congestion in the bed units that will occur. We define optimal admission rules for various critical care units in a hospital network based on each patient’s mortality risk. While maintaining a certain service level on patient blocking, the new integrated models generate optimally monotonic admission thresholds for a network including multiple IMCs and multiple Intensive Care Units (ICUs).
      The improved operational decision support methods developed meet the need to (1) reduce costly overtime and staff turnover while better utilizing physicians, nurses/staff, a room/equipment resource, (2) improve access to care (reduced time to enter a trial and fewer visits that violate the visit interval requirements of the protocol), (3) reduced wait to get an appointment, and (4) better decision support for the proper integration of services, such as clinical research studies added on to clinical care.
      This work forges new methods for data-driven approaches to patient flow modeling, resource allocation (appointment setting, which nurses or room types to use for a specific appointment from among the available options, planning for staffing levels, and the complexities surrounding patient recruitment and starting or ending clinical trials). The research included (1) developing probabilistic models that are data driven and tractably characterize the pathways the patients take through beds or services to receive healthcare, (2) the integration of optimization tools such as mixed integer programming with stochastic models of the system that incorporate queueing and blocking effects. This work represents a breakthrough in developing approximations that allow this type of queueing network control problem to be solved via math programming methods, which easily allow many important constraints to be incorporated and make the models easier to solve computationally.
      This work benefited from the collaborative efforts of multiple faculty, physicians, and medical staff (much of that unbudgeted) as it generated a cross-cutting collaborative effort involving multiple areas of three hospitals (University of Michigan, St. Joseph Mercy Hospital, and the Mayo Clinic).

  6. National Institutes of Health, Michigan Institute for Clinical and Health Research (MICHR) Renewal, $48,788,667. June 27, 2012 – May 31, 2017. PI: Thomas P. Shanley. My share: $6,750. For “MICHR Van Oyen Evaluations Faculty” Project/Grant # F030935 – Shortcode #070567. Supports 0 students.
  7. National Science Foundation,Forecasting and Control Methodology for Monitoring of Chronic Diseases,” $280,000. May 1, 2012 – Apr. 30, 2014. PI: M. Lavieri. Co-PIs: Van Oyen and J.D. Stein (Kellogg Eye Inst.). My share: $119,412. Supports 1 Ph.D. student and 1 undergraduate in summer in total.
    • The research objective of this award is to develop broadly applicable methods to improve monitoring and management of patients with chronic diseases. The systems engineering based modeling methodology will generate effective predictions of disease progression over time. Further, they will be integrated with real-time feedback-driven forecasting and control/optimization algorithms to help clinicians determine the interval of time until a particular chronic disease patient should be monitored next or an intervention should be considered by a physician. The research links population-based knowledge to patient-specific information measurements taken sequentially to determine the optimum monitoring intervals. Preliminary research indicates that linear Gaussian system models are effective for modeling progression, but existing theory will be extended to include controlled observations that optimize the tradeoff between intervals that are too short or too long. By using data to generate partially observable state space models, and by using higher dimensional state spaces, we bring a new perspective to this problem. Our work advanced the theory of “measurement subsystems control” and used it to recommend the degree of intervention as well as the costly measurements to take so as to optimize the tradeoffs over a future horizon through LQG control.   This work indicates to clinicians when glaucoma progression has occurred and how chronic diseases such as glaucoma are likely to progress. In addition, our results will provide a recommendation on when to next monitor the patient. Such knowledge will improve the health outcomes of the population and also result in cost containment.
  8. National Science Foundation, “Hospital Systems Occupancy Prediction and Control to Increase Access, Smooth Provider Workload, and Reduce Cost,” $239,708. Jun. 1, 2011 – May 31, 2014 PI: M.P. Van Oyen. Supports 1 Ph.D. student and 1 undergraduate in summer.
    • This research targets the modeling and optimization of flow in hospitals, other healthcare settings, and service sector operations more generally. The primary goal of the research was to generate new methodologies for mathematically quantifying how patients move through healthcare institutions (e.g., how they spend time in various hospital units during their entire stay). The core idea of this patient flow modeling is to create the statistical patterns for categories/types of patients that are useful for understanding how the healthcare resources/capacities are used and how appointment scheduling (referred to as admissions scheduling for a hospital) can be managed in a better way. That is, mathematical models that can effectively predict the service of healthcare patients to provide a means to optimize the upfront decisions of how many non-emergency patients of each category to allow into the hospital so that it functions better. Both elective and emergency patient arrivals were incorporated into unified models of the census in each unit/ward of the hospital on each day of the week so that the outcomes under various admission or management policies could be predicted.
      The following are examples of the types of waste or dysfunction that these new methods try to improve: We seek to limit the amount of waiting time in the Emergency Room, which is called boarding when the patient must wait hours for an inpatient bed to become available in the hospital. There is usually a most desirable bed unit/ward to which inpatients should go. When the desired ward is full patients are placed “off-unit”, and we seek to limit this because it reduces the quality of care. Proper nurse staffing levels and appropriate nurse daily assignments are harder to achieve and/or more costly when the census of the hospital is highly variable. Further, poor patient flow management leads to ambulance diversions (when the emergency department becomes saturated) and operational chaos. Part of the reason for the potential to improve these issues is that hospitals often lack effective enterprise level strategic planning of bed and care resources. Many hospitals are significantly more highly utilized during the middle of each weak, but Mondays, Tuesdays, and Fridays tend to have useable bed capacity that is wasted. Furthermore, there are many other flow-related issues that can all be modeled, analyzed, and improved, such as: reducing bed block, reducing the census variability over time, reducing the number of elective surgeries cancelled for lack of beds, and the number of patients placed off unit.
      By creating new methods to analyze and improve the functioning of the hospital in terms of the logistical flow of patients, there is great potential to reduce the cost of services, increase the number of services that can be provided, increase patient satisfaction (e.g., through reduced waits and fewer off-unit placements), and improve the quality of care provided. Our approach addresses hospitals’ internal costs and resource utilization, thereby addressing cost containment while attending to multiple other important dimensions as well.
      Key performance objectives included: reduced waits for patient access, reduced variation in patient flows, and reduction in the number of elective procedures/services that are canceled for lack of capacity (which is derived from planning and a reduction in flow variance).
      This research develops new analytical models of controlled hospital census that can, for the first time, be incorporated into a Mixed Integer Programming model to optimally solve the strategic planning/scheduling problem. We formulated a new Poisson-Arrival-Location Model (PALM) based on an innovative stochastic location process that we developed, the “Patient Temporal Resource Needs” model. We further extend the PALM approach to the class of deterministic controlled-arrival-location models and develop linearizing approximation to stochastic blocking metrics. This work provides the theoretical foundations for an efficient scheduled admissions planning system. The results of our analysis, validation, and case study are consistent with the idea that (1) smooth census levels do in fact improve blocking and throughput performance, and (2) elective admission scheduling can be used to smooth census while constraining the amount of planned off-ward placement across the network of hospital resources.
      This work forges new ground in developing data-driven approaches to patient flow modeling, resource allocation, operational flexibility, emergency department design for improved flow, and related challenges.
      This research project was an important part of the doctoral dissertations of three students who have now received their degrees. Multiple masters and undergraduate level students were involved in the performance of this work. The results of this work have been incorporated into university courses, and the results have been shared broadly, including through the internet, and published to increase the broader impact. It benefited from the collaborative efforts of engineering faculty, physicians, and medical staff as a cross-cutting collaborative effort.
  9. Veteran’s Administration –CASE interagency personnel agreement (IPA), “Development of Capacity Models and Analysis Tools for the National Fee-based Services Redesign Program” Oct. 1, 2009 – May 31, 2010. PI: M. Van Oyen. Interagency Personnel Agreement with VA for 16% of M. Van Oyen and 60% of post-doc S. AhmadBeygi.
  10. of Veterans Affairs (VA), “VA-CASE: VISN11 VA Center for Applied Systems Engineering” A Multi-organization Veterans Engineering Resource Center (VERC) consisting of five VA hospitals and U-M Ann Arbor, Dearborn, and Flint; Purdue; Wayne State). June 1, 2009-Aug 31, 2011; $1,500,000 (with no specific budget portion assigned to any individuals). Role: served as the CoE lead for over a year of proposal preparation and development, including preliminary work performed with Ph.D. student J. Helm and post-doc S. AhmadBeygi. Establishment of VERC has yielded research and support for various CoE faculty, the hiring of multiple IOE graduates, and an on-going stream of student projects (through 2021 at least).
  11. Office of Naval Research, “Development and Testing of a Hybrid Agent Approach for Set-Based Conceptual Ship Design through the Use of a Type-2 Fuzzy Logic Agent to Facilitate Communications and Negotiation,” $317,483. June 1, 2007- May 31, 2010; PI: D.J. Singer; I had a subaccount.
  12. National Science Foundation, “Collaborative Research: A Design Methodology for Operational Flexibility,” $300,000. Apr. 1, 2005 – Mar. 31, 2010 co-PI’s: M.P. Van Oyen and S.M.R. Iravani (Northwestern Univ.), My Share $150,000. Supported 2 Ph.D. students. There was an additional REU Award subsequently that involved students from Gordon College.
    • Our research has studied different mechanisms to enhance operational flexibility in a variety of ways.

      – Design of efficient structures for systems with multifunctional workers (i.e., cross-trained workers, multifunctional machines, multi-product factories).
      – Developing optimal or suboptimal control policies for worker scheduling/production control and inventory control policies in systems with flexible production resources.
      – Finding robust and flexible strategies to deal with disruption in production resources.
      – Identifying how dual-sourcing (a form of operational flexibility) and financial hedging can overlap or complement each other in providing value to a firm.
      – Development of a decentralized mechanism for a fundamental problem of supply chain coordination with arbitrary numbers of products, manufacturers, and suppliers possessing individual utility functions and production capabilities. This model emphasized asymmetric information and the model had sufficient generality to model flexible suppliers. A messaging scheme and an allocation mechanism was designed and proven to achieve the centralized optimal solution.

    • Selected Findings:

      – Design of Flexible Cross-Training Structures in Parallel Service Centers:
      For parallel service operations such as call center environments, we built upon work funded in a prior NSF grant which quantified a concept we term ‘structural flexibility’ that was based on solving a max-flow algorithm for a deterministic model of the queueing network. In this grant, we brought to fruition two different approach to the problem. (1) The first new method borrows the idea of small world networks from social sciences and develops a corresponding work sharing network for which the average (shortest) path length is calculated. It is a deterministic method to detect the more flexible agent cross-training structures among alternative structures. (2). In addition, we also extended the Structural Flexibility method to Capacitated Flexibility method that deals with systems in which agents have different capacities (different speeds). Both methods result in a single number that is used as a metric to compare two different cross-training structures and choose the most flexible one. Our numerical study has shown that these metrics are very accurate in identify flexible structures that result in shorter customer average waiting times in call centers. This research made the strategic method of Structural Flexibility more practical and suitable for tactical applications and design at a more detailed level.

      – Integrated Inventory and Process Flexibility: We have studied the interaction between the process flexibility (i.e., a production process capable of producing more than one item) and inventory flexibility (i.e., the ability to substitute one product for another). Using a Markov Decision
      Process we characterized the structure of the optimal integrated production and inventory substitution policies and showed that it has a very complex structure, even in a simple produce-to-stock system. We, therefore, developed a simple heuristic that is nearly as cost-effective as the complex optimal policy. We have also provided a series of managerial insights into when process
      or inventory flexibility is most beneficial.

      – Strategic Risk in Supply Chains with Supplier Disruption. We have studied how manufacturers can increase their flexibility to respond effectively to disruption in their supplies. We have developed analytical models to find the optimal strategies for pre-disruption and post-disruption periods. We have identified situations under which the manufacturers are in higher strategic risk (i.e., they lose a larger portion of their market share) and we analyzed policies that mitigate the impact of the disruption.

      – Dual Sourcing and Financial Protection. Suppliers are at risk of disruption and therefore, the manufacturer or service provider is also at risk of disruption. Operationally, the manufacturer may choose to source from two (or more) suppliers instead of one. Financially, the manufacturer may choose to purchase financial protection to mitigate the risk of the disruption of the incoming financial flows. The work illuminates the way in which operational and financial hedging/flexibility can overlap or complement each other in providing value to a firm.

  13. National Science Foundation, Student Support Award for Operational Flexibility. $44,292. Sept. 1, 2007 -Aug. 31, 2009. PI: M.P. Van Oyen. Supported 1 Ph.D. student.
    • To emphasize one part of this work, We studied alternative dynamic assignment policies in an information technology (IT) service delivery environment. The goal of these policies is to find the most efficient assignment of service requests to cross-trained agents in a large-scale network. We also present a heuristic algorithm that assigns a priority-based assignment index to each service request. This novel algorithm incorporates factors such as variability in agents’ capabilities, uncertainty in request inter-arrival times and complex service level agreements (SLA) and penalties for non-performance. We validated the effectiveness of our proposed assignment algorithm using real world data from an IT service environment. We finally discuss how the results of this simulation can help improve terms of service level contracts as well as agent training programs.

      In parallel to the above, we developed an MDP model to capture the essential feature of the IBM-motivated problem of dynamic control of heterogeneous flexible servers working under a contract to provide a specific service level (with penalties for non-compliance). The work is theoretically challenging and we believe we have a rigorous structural characterization of an optimal policy. In addition, we have developed an effective heuristic.

  14. Loyola University Chicago competitive Summer Research Grant, “Operational Flexibility for Design;’’ $ June and July of 2005. PI: M.P. Van Oyen.
  15. National Science Foundation, “Collaborative Research: Robust Strategies for Cross-training Call Center Agents – Taxonomy, Models, and Analysis;” $374,878. Aug. 15, 2001-Aug. 14, 2004; co-PI’s: M.P. Van Oyen and S.M.R. Iravani, My Share $185,445. Supported 2 graduate students.
    • Over the past two decades, the typical American business has vigorously worked to implement bold new visions for their operations. This has resulted in a broad-based shift from workers minimally trained for one task to cross-trained workforces with multiple tasks and dynamic worksharing. A vivid illustration of this change can be found in call centers, a large service industry employing roughly 3-4 million Americans, growing at about 10% annually, according to Data Monitor. Critical emergency services such as 911, police, ambulance, and fire dispatching depend on call centers, so much more than convenience and profit are at stake. In general, competitive marketplace pressures, increasing customer service expectations, and the recent advent of skills-based routing technologies have caused a dramatic shift toward cross-trained customer service representatives (CSR’s). That is, multiple call types/products are now being handled during a shift by the same agent.
      Existing research efforts focus on issues such as staffing levels, skills-based routing, and CSR scheduling. Because skills-based call (and data communications) routing systems have recently become standard equipment, there is a pressing need to develop strategies for CSR cross-training. The PI’s discussions with industrial call center managers and software solution providers also suggest that strategies for cross-training and call/agent assignment are ripe for research and promise a significant step forward in call center managerial practice and performance.
      The objectives of this research are (1) to develop a rigorous, scientific framework and supporting models for setting a strategy for CSR cross-training in a call centers, and (2) to provide tactical-level policies that enable call centers to harness the flexible capacity made available through agent cross- training combined with skills-based routing technology. This approach considers the dependence of system congestion upon the cross-training approach taken, cross-training’s effects on CSRs, and the quality of service provided to the customer.
      This research will construct a detailed, conceptual classification scheme for call center environ- ments that identifies key characteristics germane to the selection of a cross training strategy. It will create and analyze a series of models that predict the performance of various cross-training patterns utilizing skills-based routing and provide insight into the factors that determine their efficacy. The primary focus of the research will be on logistical efficiency (e.g., optimizing performance in terms of throughput, response time, or call abandonment rate for a given workforce size). This effort will develop original, detailed strategies for the use of cross-training to achieve highly effective skill patterns in the labor force as well as synergistic call/agent routing policies. The analysis will use tools that include queueing theory, Markov Decision Processes (MDP’s), and discrete event systems theory and simulation. Historically, call centers have relied on queueing models (especially single-class Erlang loss models), but the project employs the MDP and discrete event simulation approach to support the development of heuristic design strategies. This research will be carried out in close collaboration with several partnering industrial organizations to ensure that the project’s results can be implemented and address issues such as motivational effects, human factors, and managerial complexity.
      The anticipated results of this research are: (1) managerial insights that greatly deepen the understanding of which systems will benefit from cross-training and a suitable strategy for im- plementation, (2) CSR cross-training strategies that are robustly effective across a wide range of call centers, (3) useful analytical models for the analysis and design of agile work systems, and (4) extension of the queueing technology base to include broad classes of systems where servers operate in new and complex ways based on their skill sets.
  16. Loyola University Chicago competitive Summer Research Grant, “Quantifying Operational Flexibility,” $8000. June and July of 2004. PI: M.P. Van Oyen. Supported faculty research effort.
  17. Loyola University Chicago competitive Summer Research Grant, “Improving Call Center Performance Through Advanced Operations Management,” $8000. May 15, 2001 – May 14, 2002. PI: M.P. Van Oyen.
  18. National Science Foundation, “Workforce Agility: Classification and Modeling,” $375,000. Jun. 1, 1998 – May 31, 2002, PI: W.J. Hopp. Co-PI: M.P. Van Oyen, My Share $185,445. Supported 2 graduate students.
    • The typical American workplace has undergone radical change in the past two decades. Traditional division-of-labor schemes have given way to a plethora of organizations that make use of cross-trained workers, flexible task assignment policies, employee decision-making empowerment, team-oriented incentive structures, and many other modern methods. Under pressure from global competition to maximize labor productivity, more and more manufacturing and service firms are striving to organize their workforces in new and more agile ways. While there are many examples of agile work systems in industry, some of which have been studied by academics, there is as yet no general structure within which to understand them. As a result, practitioners have little basis for identifying workforce policies that are well-suited to specific manufacturing/service environments.

      This research project is directed at developing a comprehensive science of agile work systems by (a) constructing a detailed classification scheme for manufacturing/service environments that identifies key characteristics germane to the selection of a workforce policy, and (b) creating and analyzing a series of models with which to predict the performance of various policies in various environments and thereby gain insight into the factors that determine their efficacy. The primary focus of the research will be on logistical efficiency (e.g., maximizing performance in terms of throughput, flowtime, or work-in-process for a given workforce size). But, because this research will be carried out in close collaboration with several industrial practitioners, real world considerations such as motivational effects, human factors, and managerial complexity, will be addressed as well.

      The anticipated impacts of this research are: (a) managerial insights that greatly deepen understanding of why and where agility in the workforce is effective, (b) a set of useful analytical models (performance predictors, quantification of opportunity, and near-optimal policies/organization schemes) to assist engineers in the analysis/design of agile work systems, and (c) extension of the queueing technology base to include broad classes of systems where machines and labor operate in new and complex ways. Technology transfer will occur through practitioner and research oriented publications, direct contact with our industrial sponsors, and our own teaching efforts to develop workforce agility insights in future engineering and management practitioners.

  19. ALCOA Science Foundation, “Control of Queueing Systems, Markov Decision Processes, and Stochastic Scheduling,” Aug. 1, 1997 – Aug. 31, 1999. PI: M.P. Van Oyen.
  20. National Science Foundation, “Stochastic Scheduling Methods for Queueing Systems,” $165,000. Sept. 1, 1995 – Aug. 31, 1999; PI: M.P. Van Oyen. My Share $165,000. Supported 1 graduate student.
  21. General Motors Foundation, “Human Assistive Devices for Vehicle Assembly,” $500,000. Jul. 1, 1995 – Jun. 30, 2000. PI: Abraham Haddad Co-PI: J. Edward Colgate, Lina L.E. Massone, Lucy Y. Pao, Michael A. Peshkin, and M.P. Van Oyen.. Primary emphasis of grant was to provide scholarships and support underrepresented students as well as to bring together a Council on Dynamic Systems and Control within NU. Supported multiple students.
  22. Northwestern University, “Numerical Optimization of Fundamental Queueing Systems with Overhead,” 1, 1994 – May 31, 1995. PI: M.P. Van Oyen.
  23. Electric Power Research Institute, “RP 3599 UCA Integrated Protection and Control WO3599-05;” $196,996. Jan. 1994 – Dec. 1994. PI: W. Premerlani. Co-PIs: R.J. Mitchell and M.P. Van Oyen. Note: I was a co-PI on the grant submission, but then left GE and the award was modified.