Decision Analytics for Complex Systems

Patrick Reed, School of Civil & Environmental Engineering


From Reed Research Group | Decision Analytics for Complex Systems

CEE 5980 Introduction to Decision Analysis

This course seeks to provide you with a framework to structure the way you think about decision situations that are complicated by uncertainty, complexity, and competing objectives. Creative problem framing and generation of decision alternatives will be pivotal for confronting highly complex and uncertain decision contexts. Modern decision making routinely requires a careful balance of economic development, environmental sustainability, and evolving uncertainties. The concepts presented build on more than 50 years of research and practice in management sciences, operations research, multiobjective decision theory, and the engineering systems design.

Course Topics • Introduction. What is decision analysis? Why do it? Elements of decisions. • Value-focused thinking. • Structuring values. Framing decisions. • Review of probability basics. • Decision trees. • Dominance. Multiobjective decisions. • Sensitivity analysis • Generating alternatives. Blocks to creativity. • Theoretical probability models. • Subjective probability assessment. • Value of information. • Risk attitudes. Utility functions. • Multiattribute utility functions. • Monte Carlo simulation.

CEE 6200 Water Resources Systems Engineering

Growing concerns about how “change” (climate, land-use, population, etc.) will strain our water resources is motivating the need for the next generation of professionals that can innovate the planning and management of these systems. Course topics build on the legacy of research in the water resources systems area and seek to provide a new generation of planners with an enhanced ability to discover and negotiate the highly uncertain tradeoffs we face in balancing the water resources demands of the future. Students will be encouraged to explore what sustainable water management means given conflicting demands from renewable energy systems, ecosystem services, expanding populations, and climate change. Students will learn to develop and apply deterministic and stochastic optimization and simulation models for aiding in water-resources planning and management. This course covers river-basin modeling, including water allocation to multiple purposes, reservoir design and operation, irrigation planning and operation, hydropower-capacity development, flow augmentation, flood control and protection, and urban water supply portfolio management. Student projects will encourage a deeper exploration of topics and tools of interest.

Course Topics • Integrated Water Resources Planning • Water Resources Systems Modeling • Frameworks for Decision Support • Methods for Evaluating Alternatives • Optimization Methods (LP, DP, Heuristics) • Dominance. Multiobjective decisions. • Sensitivity analysis • Uncertainty Analysis • Monte Carlo simulation • River basin planning • Urban water supply • Climate change • Planning under Deep Uncertainty

CEE 6660 or SYSEN 6410 Multiobjective Systems Engineering Under Uncertainty

Computing power is cheap, optimization algorithms have advanced, visualizations have gotten more sophisticated, and many systems optimization frameworks have found their way into commercial use, yet for nearly 25 years we have continued to formulate and solve complex engineered systems design problems largely the same way. Many of the innovations discussed above have occurred in isolation, and the design of complex engineered systems remains plagued by neglected uncertainties and narrow definitions of optimality that fail to encompass critical performance tradeoffs. This course will use programming case studies and projects drawn from the disciplines of participating students to introduce emerging multiobjective design tools that enhance systems engineering design under uncertainty. Beyond the focus on multiobjective optimization, the course will also expose students to relevant research related to constructive decision-aiding theory, decision-biases identified in the behavioral economics literature related to risk, and recent advances in visual analytics that have been demonstrated to enhance design processes. A core goal of this course is to help students integrate human intelligence and computational power to effectively explore design hypotheses, discover critical system tradeoffs, and facilitate robust design decisions.

Course Topics • Historical origins of multiobjective decision theories • A Priori, A Posteriori, & Interactive Preference Elicitation • Dominance, Pareto Optimality Concepts • Simulation—Optimization in Design • Monte Carlo Simulation & Uncertainty Analysis • “Many-Objective” Problem Framing • Evolutionary Multiobjective Optimization Algorithms & Search Diagnostics • Multiobjective Optimization Under Well-Characterized Uncertainty • Parallel Computing Support Tools • Interactive Visual Design Analytics • Sensitivity Analysis • Deep or Knightian Uncertainty • Evaluating Satisficing versus Regret-based Robustness Given Deep Uncertainties • Robust decision making • Information Gap Analysis

Personal tools