A Comprehensive Review of Digital Twin - Part 2: Roles of

Uncertainty Quanti cation and Optimization, a Battery

Digital Twin, and Perspectives

更多

Adam Thelen1, Xiaoge Zhang2, Olga Fink3, Yan Lu4, Sayan Ghosh5, Byeng D.

Youn6, Michael D. Todd7, Sankaran Mahadevan8, Chao Hu1* and Zhen Hu9*

1Department of Mechanical Engineering, Iowa State University, Ames, 50011, IA, USA.

2Department of Industrial and Systems Engineering, The Hong Kong Polytechnic

University, Kowloon, Hong Kong.

3Intelligent Maintenance and Operations Systems, Swiss Federal Institute of Technology

Lausanne, Lausanne, 12309, NY, Switzerland.

4Information Modeling and Testing Group, National Institute of Standards and

Technology, Gaithersburg, 20877, MD, USA.

5Probabilistic Design and Optimization group, GE Research, Niskayuna, 12309, NY, USA.

6Department of Mechanical Engineering, Seoul National University, Gwanak-gu, 151-742,

Seoul, Republic of Korea.

7Department of Structural Engineering, University of California, San Diego, La Jolla,

92093, CA, USA.

8Department of Civil and Environmental Engineering, Vanderbilt University, Nashville,

37235, TN, USA.

9Department of Industrial and Manufacturing Systems Engineering, University of

Michigan-Dearborn, Dearborn, 48128, MI, USA.

收起

Causality and Deep Learning:

Synergies, Challenges and the Future

Yoshua Bengio, Rosemary Ke

Tutorial:

Quantifying Predictive Uncertainty Without

Distributional Assumptions Via Conformal Prediction

更多

Rina Foygel Barber

http://rinafb.github.io/

收起

OPTIMIZATION

UNDER UNCERTAINTY

LECTURE NOTES

更多

2001

R. T. Rockafellar

University of Washington

收起

Risk and Uncertainty in Soviet Deliberations About War

by Benjamin S. Lambeth

Surveys the elements of risk, uncertainty, and unpredictability that might moderate Soviet behavior and undermine the confidence with which Soviet decisionmakers would consider entering into a major military engagement with the United States. Although the report does not question the substantial threat implications of Soviet force improvements that have been under way in recent years, it does describe certain realities of Soviet style and leadership concern about possible Soviet military inadequacies that make the more ominous features of Soviet doctrine and force development appear somewhat less alarming. The analysis is based on a combination of evidence suggested by past Soviet crisis behavior, information offered in Soviet literature concerning troop management and training, and inferences from known or suspected Soviet political practices, organizational characteristics, and operational concerns.

Operational Unpredictability and Deterrence

Evaluating Options for Complicating Adversary Decisionmaking

by Miranda Priebe, Angela O'Mahony, Bryan Frederick, Alyssa Demus, Bonny Lin, Michelle Grisé, Derek Eaton, Abby Doll

更多

The 2018 U.S. National Defense Strategy instructed the U.S. military to become more operationally unpredictable and suggested that doing so would help the United States deter attacks on U.S. partners. The authors of this report propose a definition of U.S. operational unpredictability—adversary uncertainty about how the United States would fight; develop four potential approaches for increasing U.S. operational unpredictability and deterring attacks on U.S. allies and partners; and assess how the four approaches could affect U.S. relations with Russia and China. They also examine two Cold War–era cases in which the United States sought to be more operationally unpredictable.

The authors find that increasing adversaries' perceptions of U.S. operational unpredictability may be possible if the United States has detailed information about their operational analysis and decisionmaking processes. The most promising way to increase U.S. operational unpredictability is to publicize new U.S. capabilities and demonstrate that they give the United States multiple options for achieving its key objectives. However, increasing U.S. operational unpredictability may be costly and, in some cases, involve negative side effects (e.g., reducing U.S. military effectiveness and increasing China's and Russia's threat perceptions). The authors recommend weighing the potential costs and effectiveness of these approaches against more traditional approaches to deterring U.S. adversaries.

Key Findings

Increasing operational unpredictability is possible but likely costly, and may have negative side effects

Increasing U.S. operational unpredictability is only one of many approaches to enhancing extended deterrence.

Limited information on Russian and Chinese intelligence, military planning, and decisionmaking processes makes it difficult to assess the potential effects of U.S. attempts to be operationally unpredictable.

The available information suggests that increasing Russian and Chinese perceptions of U.S. operational unpredictability may be possible, but also costly.

The most promising way to increase U.S. operational unpredictability is to publicize new U.S. capabilities and demonstrate that they give the United States multiple options for achieving its key objectives.

This approach may enhance extended deterrence if each U.S. way of fighting requires different and costly adversary counters.

This approach may have negative side effects, such as reducing U.S. readiness and increasing China's and Russia's threat perceptions.

Recommendations

Compare the use of operational unpredictability with alternative approaches to deterring U.S. adversaries.

Develop a clear logic linking activities intended to enhance U.S. operational unpredictability to desired outcomes, and consider potential trade-offs.

Review existing intelligence and consider increased collection on Russia's and China's intelligence, military planning, and decisionmaking processes.

Continue initiatives on Army and U.S. Department of Defense flexibility and agility, which may also increase U.S. operational unpredictability.

收起

Air Force Capability Development Planning

Analytical Methods to Support Investment Decisions

by James A. Leftwich, Debra Knopman, Jordan R. Fischbach, Michael J. D. Vermeer, Kristin Van Abel, Nidhi Kalra

更多

In view of uncertainties about the future threat environment, trajectories of technological development, and shifting budgeting priorities, RAND Project AIR FORCE examined analytical methods that would best guide the recently established Air Force Warfighting Integration Capability in capability development planning (CDP), as well as concept development and future force design. In their review, researchers found that specific methods of decisionmaking under deep uncertainty (DMDU) can provide the most suitable means for arriving at solutions that are flexible, adaptable, and robust and guiding investment pathways and modernization efforts.

The method highlighted, Robust Decisionmaking, rests on a simple concept. Rather than using models and data to assess policies under a single set of assumptions, RDM runs models over hundreds to thousands of different sets of assumptions about the problem space with the aim of understanding how plans perform under many plausible conditions. Each of the four steps of RDM — decision-framing, case generation, vulnerability assessment and scenario discovery, and trade-off analysis — feeds into the next, providing stakeholders and decisionmakers with a more informed tradespace. By exposing vulnerabilities of different options under different scenarios, RDM can enable stakeholders to engage in a rich dialogue about which risks or vulnerabilities are acceptable, as well as to review assumptions made in framing the problem and make adjustments as needed.

Key Findings

Compared with existing methods currently in use for CDP, DMDU methods — and specifically the RDM paradigm — could support investment decisionmaking in situations when evidence is lacking to quantify the probability of future threats and risks and in the face of deep uncertainties.

A workshop in which RAND experts and Air Force stakeholders explored ways to frame a decision problem revealed the layers of complexity embedded within the CDP challenge to produce decision-quality analysis to support design and planning choices.

Workshop participants identified technical limitations to implementing RDM, including availability of models and their fidelity.

Participants also noted that the practicalities of developing competencies in a new approach for which the capabilities may need to be built rather than simply adapted will be a challenge. A related issue is the adequacy of internal staffing and capacity to support CDP.

RDM and similar methods represent a fundamental shift in problem-solving from prediction-then-act analysis and decisionmaking to a risk management approach of seeking robust solutions across a range of future threat environments and other uncertainties. The key output of this type of approach is not a single answer but rather trade-off curves or other visualizations that compare performance of alternative strategies or portfolios of options with the status quo or other baseline in terms of metrics that reflect what is most important to the Air Force.

Recommendations

A first step in strategic developing planning and experimentation would be to choose a timely and relevant investment decision problem and demonstrate the robustness of a number of alternative investment portfolios using a mostly qualitative approach that would alleviate the need for sophisticated — or nonexistent — models capable of encompassing a range of emerging technologies, such as quantum computing, artificial intelligence, hypersonics, or directed energy. The analysis should consider performance under a wide range of future conditions and visualize trade-offs among these alternatives across the performance metrics relevant to Air Force decisionmakers.

To be workable, a test application will need to be amenable to analysis. Models should be understandable and appropriate for exploring broad-ranging uncertainty analysis; in the absence of numerical models, conceptual models may be used.

In the test application, there should be basic access to data that could be used to establish performance under baseline or status-quo concepts of operation and existing capabilities.

In the test application, classification levels should be managed to enable serious analysis and reveal key developments underway but not overly restrict the necessary audience of key analysts and decisionmakers.

Air Force stakeholders should be willing to engage in an iterative process in developing a decision-framing matrix, including the identification of performance metrics, key uncertainties, modeling tools, and investment/divestment strategies (levers).

The problem can be scoped to accommodate a qualitative analysis within 8–12 months of initiation, including all relevant data collection, modeling, analysis, visualization, and interactive review of and response to results.

Table of Contents

Chapter One

Introduction

Chapter Two

Capability Development Planning in the Air Force

Chapter Three

Current Assessment and Aspirational View of Capability Development Planning

Chapter Four

Analytic Methods for Capability Development Planning

Chapter Five

Decision-Framing Exercise

Chapter Six

Recommendations and Next Steps

Appendix

Background Information on Capability Development Planning and Strategy, Planning, and Programming Process

收起

Technology and Work Worth Doing (Not Jobs) Post-AlphaGo

A World-Building Workshop on the Future of Artificial Intelligence

by Ann Pendleton-Jullian, Robert J. Lempert

更多

There has been extensive discussion of artificial intelligence's (AI's) impacts, good and bad, but much less on how AI might be used to shape a new world in which we would like to live. In June 2018, in Santa Monica, California, the Pardee RAND Graduate School and the RAND Frederick S. Pardee Center for Longer Range Global Policy and the Future Human Condition convened a workshop called "The Future of Technology and Work (Not Jobs) Post-AlphaGo." This workshop gathered a small group of innovative thinkers to engage in an approach called large-scale speculative design and its companion methodology, world-building, to sketch desirable, AI-enabled future worlds. Large-scale speculative design aims to avoid the inertia of the present by helping us to imagine possible outcomes that may be different from what is already set in motion. World-building is characterized by imagining a few alternative and desirable future "what if" propositions and then building them out in detail across many relevant domains to understand their implications, uncover unintended interactions, and experiment with causal chains. This workshop was intended both to prototype a new methodology for speculating on an aspect of the future human condition and to initiate a line of inquiry around the topic from this different approach. This report summarizes the framing of the topic, the process that was employed, the three scenarios and initial world-building exercises, and the ensuing conversation among the workshop participants. It finishes by outlining suggested next steps.

Table of Contents

Chapter One

Introduction

Chapter Two

Project Framing

Chapter Three

World-Building Methodology

Chapter Four

Workshop Description

Chapter Five

Workshop Results

Chapter Six

Next Steps

Appendix A

Workshop Participants

Appendix B

Workshop Agenda

Appendix C

Some Participants' Post-Workshop Reflections

收起