Institute for Systems Research Technical Reports

Permanent URI for this collectionhttp://hdl.handle.net/1903/4376

This archive contains a collection of reports generated by the faculty and students of the Institute for Systems Research (ISR), a permanent, interdisciplinary research unit in the A. James Clark School of Engineering at the University of Maryland. ISR-based projects are conducted through partnerships with industry and government, bringing together faculty and students from multiple academic departments and colleges across the university.

Browse

Search Results

Now showing 1 - 10 of 10

Data-driven Metareasoning for Collaborative Autonomous Systems
(2020-01) Herrmann, Jeffrey
When coordinating their actions to accomplish a mission, the agents in a multi-agent system may use a collaboration algorithm to determine which agent performs which task. This paper describes a novel data-driven metareasoning approach that generates a metareasoning policy that the agents can use whenever they must collaborate to assign tasks. This metareasoning approach collects data about the performance of the algorithms at many decision points and uses this data to train a set of surrogate models that can estimate the expected performance of different algorithms. This yields a metareasoning policy that, based on the current state of the system, estimated the algorithms’ expected performance and chose the best one. For a ship protection scenario, computational results show that one version of the metareasoning policy performed as well as the best component algorithm but required less computational effort. The proposed data-driven metareasoning approach could be a promising tool for developing policies to control multi-agent autonomous systems.
A Framework for Design Theory and Methodology Research
(2016-04) Herrmann, Jeffrey
The scholarly study of design continues to develop new knowledge through a variety of approaches. Some researchers examine how designers work, and many develop new methods to help designers do design tasks. Studying design is complex for many reasons. There are many domains in which design occurs, including all of the disciplines of engineering, architecture, and other fields. More significantly, humans design, and human behavior can be difficult to understand. Designers sometimes work alone and sometimes in a group or team. Designers experience design work in multiple ways. Design researchers have been exploring many different aspects of design and experimenting with many different approaches and generating a variety of different design theories. The focus on exploration, however, has meant that there has been less emphasis on exploiting previous research and creating an organized body of knowledge. Building a unified body of knowledge is a long-term challenge. This paper describes a proposed framework for design theory and methodology research. This framework, which is based on ideas from education research, does not specify specific topics or methodologies. Instead, it describes six different research types: (1) Foundational Research, (2) Early-Stage or Exploratory Research, (3) Design and Development Research, (4) Efficacy Research, (5) Effectiveness Research, and (6) Scale-up Research. Illustrating these types are examples based on a table design example. The paper explains how these six research types are related to each other and how, collectively, they serve to generate valid knowledge about design. The research types follow a logical sequence in which researchers develop basic knowledge, create design methods, and test design methods. Although the framework numbers the research types following this natural progression, it does not insist that researchers do or should work by rigidly following this sequence. These research types actually form a cycle of research that iterates through three “phases”: description, explanation, and testing. In this cycle, researchers observe and describe a phenomenon, develop theories to explain the phenomenon and its interactions and effects, and test that theory against the phenomenon, and then, based on the results, refine their descriptions, revise their theories, and conduct more testing. Over time, the description of the phenomenon is improved (e.g., made more precise or more general), better explanations (theories) are found, and additional testing further demonstrates their correctness (or indicates their limitations). The proposed framework can show how different research studies are related to each other because they are the same research type or they fit into the progress of a design theory or the development of a design method. Thus, the proposed framework, while not a theory of design, can help researchers respond to the challenges of coordinating the different types of research needed to create useful design theories and build a unified body of knowledge. Future work is needed to analyze, test, and refine this framework so that it becomes truly useful to the design research community.
Modeling Separations of Plant Layout Problems
(2014-08) Herrmann, Jeffrey
This paper describes the results of a simulation study that evaluated the performance of different separations of the plant layout problem solved by bounded rational decision-makers. Seven problem instances from the literature were studied. We simulated the solution of a problem by a bounded rational decision-maker as a random search over the solution space. The problem was separated by identifying “subsets” of adjacent locations. The subset assignment problem partitioned the departments into subsets corresponding to these subsets of locations. Then, the subset layout problem assigned the locations in the subset to the departments. We considered separations with 2, 3, and 4 subsets. We also considered separations that first aggregated the departments before assigning them to subsets of locations. The results showed that separating the problem can lead to better solutions than solving the problem all-at-once, but some separations lead to worse solutions. Maximizing the flow inside the subsets generated better solutions than maximizing the adjacency of the departments inside the subsets. When fewer subsets are used, minimizing the cost inside each subset generated better solutions than minimizing the total cost. These results show that the quality of the solutions created by a design process is influenced by the choice of subproblems that make up the design process.
Predicting the Performance of Teams of Bounded Rational Decision-makers Using a Markov Chain Model
(2013-08) Herrmann, Jeffrey
In practice, when faced with a complex optimization problem, teams of human decision-makers often separate it into subproblems and then solve each subproblem instead of tackling the complete problem. It would be useful to know the conditions in which separating the problem is the superior approach and how the subproblems should be assigned to members of the teams. This paper describes a mathematical model of a search that represents a bounded rational decision-maker (“agent”) solving a generic optimization problem. The agent’s search can be modeled as a discrete-time Markov chain, which allows one to calculate the probability distribution of the value of the solution that the agent will find. We compared the distributions generated by the model to the distribution of results from searches of solutions to traveling salesman problems. Using this model, we evaluated the performance of two- and three-agent teams who used different solution approaches to solve generic optimization problems. In the “all-at-once” approach, the agents collaborate to search the entire set of solutions in a sequential manner: the next agent begins where the previous agent stopped. In the “separation” approach, the agents separate the problem into two subproblems: (1) find the best set of solutions, and (2) find the best solution in that set. The results show that teams found better solutions using separation when high-value solutions are less likely. Using the all-at-once approach yielded better results when the values were uniformly distributed. The optimal assignment of subproblems to teams also depended upon the distribution of values in the solution space.
Separating the Searches of Bounded Rational Decision-Makers
(2013-06) Herrmann, Jeffrey
In practice, when faced with a complex optimization problem, human decision-makers often separate it into subproblems and then solve each subproblem instead of tackling the complete problem. This paper describes a study that simulated small teams of bounded rational decision-makers (“agents”) who try different approaches to solve optimization problems. In the “all-at-once” approaches, the agents collaborate to search the entire set of solutions in a sequential manner: each agent begins where the previous agent stopped. In other approaches, the agents separate the problem into subproblems, and each agent solves a different subproblem. Finally, in the hybrid approaches, the agents separate the problem but two agents will collaborate to solve one subproblem while another agent solves a different subproblem. In some cases, the subproblems are solved in parallel; in others, the subproblems are solved sequentially. The results show that the teams generally found better solutions when they separated the problem.
Scheduling Perfectly Periodic Services Quickly with Aggregation
(2013-03) Herrmann, Jeffrey
The problem of scheduling periodic services that have different period lengths seeks to find a schedule in which the workload is nearly the same in every time unit. A time unit’s workload is the sum of the workloads of the services scheduled for that time unit. A level workload minimizes the variability in the resources required and simplifies capacity and production planning. This paper considers the problem in which the schedule for each service must be perfectly periodic, and the schedule length is a multiple of the services’ period lengths. The objective is to minimize the maximum workload. The problem is strongly NP-hard, but there exist heuristics that perform well when the number of services is large. Because many services will have the same period length, we developed a new aggregation approach that separates the problem into subproblems for each period length, uses the subproblem solutions to form aggregate services, schedules these, and then creates a solution to the original instance. We also developed an approach that separates the problem into subproblems based on a partition of the period lengths. Computational experiments show that using aggregation generates high-quality solutions and reduces computational effort. The quality of the partition approach depended upon the partition used.
Optimizing Urgent Material Delivery by Maximizing Inventory Slack
(2012-06-12) Montjoy, Adam; Herrmann, Jeffrey
Motivated by the need to create plans for delivering medication quickly during a public health emergency, this paper formulates the Inventory Slack Routing Problem. The planning goal is to deliver material as early as possible to demand sites from a central depot at which material arrives over time. The objective function is to maximize the minimum slack of the deliveries so that all demand sites are treated equitably. This paper presents and analyzes the problem, discusses solution techniques, and discusses the results of computational experiments used to compare the solution techniques. Although motivated by planning for public health emergencies, this work is also applicable to other settings in which material must be delivered quickly to multiple facilities that rely upon the material to operate.
Optimally Allocating MedKits to Defend Urban Areas from Anthrax Attacks
(2011-07) Herrmann, Jeffrey
The deliberate release of aerosolized anthrax spores in a large city will expose thousands to this deadly disease. Although state and local health departments have developed contingency plans for promptly opening points of dispensing (PODs) and distributing antibiotics to those exposed after an attack is detected, other risk mitigation strategies have been proposed. This study focuses on the pre-event placement of pharmaceuticals in individual households for use only as directed by public health authorities. The pre-deployed medications are commonly known as “MedKits.” This paper considers the problem of a defender who wishes to minimize the expected fatalities of an anthrax attack by allocating a limited number of MedKits to various urban areas. Under the condition that the attacker wishes to maximize the expected fatalities, the defender’s optimal policy is to keep all of the potential targets equally attractive. The paper presents a methodology for finding this optimal policy. The paper considers a specific example using ten urban areas in the United States and compares the optimal policies with those in which the MedKit allocations are proportional to population. The approach can be adapted to consider a wide range of scenarios and local factors to help public health officials manage the risk of an anthrax attack. Having good solutions to this problem should be valuable to public health officials who are considering how to pre-deploy MedKits.
Predicting the Impact of Placing Pre-event Pharmaceuticals for Anthrax
(2011-01) Houck, Michelle; Herrmann, Jeffrey
Finding feasible strategies to distribute antibiotics quickly to the general public in response to an anthrax attack remains a difficult challenge. Among the proposed strategies is the pre-event placement of pharmaceuticals in individual households for use only as directed by public health authorities. These medications (known as “MedKits”) would allow many exposed persons to begin treatment quickly while reducing the number who visit on points of dispensing, the primary distribution strategy. This paper describes a model that estimates the expected number of deaths in an anthrax attack by modeling the logistics of the response and the use of MedKits. The results show that increasing the number of MedKits distributed can reduce the expected number of deaths. When the population has more potential exposures, deploying MedKits is more effective. The MedKits reduce the number of potential exposures who seek prophylaxis, which allows those truly exposed (but without MedKits) to receive medication sooner, which saves lives. Beyond the scenarios considered here, the ability to predict this benefit in other scenarios will be valuable to public health officials who are considering this option.
Solving Continuous Replenishment Inventory Routing Problems with Route Duration Bounds
(2010-01) Herrmann, Jeffrey; Fomundam, Samuel
In a public health emergency, resupplying points of dispensing (PODs) with the smallest number of vehicles is an important problem in mass dispensing operations. To solve this problem, this paper describes the Continuous Replenishment Inventory Routing Problem (CRIRP) and presents heuristics for finding feasible solutions when the duration of vehicle routes cannot exceed a given bound. This paper describes a special case of the CRIRP that is equivalent to the bin-packing problem. For the general problem, the paper presents an aggregation approach that combines low-demand sites that are close to one another. We discuss the results of computational tests used to assess the quality and computational effort of the heuristics and the aggregation approach.

Institute for Systems Research Technical Reports

Browse

Filters

Settings

Sort By

Results per page

Search Results