Two Vectors Towards a Grand Unified Theory of Systems Engineering usc851 Dr. George Friedman Associate Director, Systems Architecting and Engineering Program University of Southern California Center for Systems and Software Engineering (CSSE) Convocation 2006 Selected Quotes: From 45+ years as a working guy: “I don’t have time to drain the swamp when I’m up to my ass in alligators” “Paralysis though analysis” “We don’t have time to do it right, but we always seem to have the time to do it again” “The data is not available and we don’t have time to collect it” (but even when it is available, it’s used incorrectly) “We have neither the time or budget to test thoroughly” Then, as 3rd president of INCOSE: “What happened to the ‘systems’ perspective? The SE process family is stovepiped!” Why GUTSE? Presently, Systems Engineering has weaknesses, including: Multi-stakeholder, multi-criteria --> incoherent, political decisions Advanced technologies, new configurations, environments --> deterministic design and modeling in uncertain world Excessive decomposition-->incomplete FFDs, block diagrams,’walls’ Insufficient academic recognition and “intellectual content” --> education vs training; core expertise not captured Two vectors for above: Decision Theory and Graph Theory Decision Theory NSF: “Design is merely a series of decisions” --> “Decision-based design” thrust --> apply the body of decision theory research to engineering GJF: Expand and quantify risk management --> balances performance, cost and schedule triad --> acknowledges probabilistic nature of technology, environment --> meta-decisions on utilities and rationality converge consensus --> decision diagrams provide team wide visibility --> encourages testing; even imperfect tests (responds to schedule, budget, quality issues of testing) Graph Theory Human cognition ~7 dimensions but Modeling needs 104 + Decomposition and re-integration is treacherous Leads to suboptimization and suppression of interactions Path to complexity management: Multi-D math modeling Let the primary complexity driver provide tools as well Constraint Theory employs bipartite graph metamodels GJ Friedman, Constraint Theory, Multi-D Math Model Mgt, Springer 2005 Addresses the “well-posed” problems of consistency, computability The “BNS” is the kernel of constraint; at the U{circuits} Easy properties of the bipartite graph --> well-posed conclusions Bipartite Graphs for Right Brain Rigor usc540 DECISION THEORY Tree-like : Decisions : Events CONSTRAINT THEORY Circuits and Trees : : Relations Variables HOW GUTSY IS GUTSE? Compared to the central reductionist dogma of science, which “believes” that intelligence can be explained from biology, biology from chemistry, chemistry from physics, and physics from cosmology, GUTSE should be easier because: we can (or should) choose from mature technologies, far from the frontiers of science, such as 4-force unification, the origin of life, the evolution of intelligence, chaos… GUTSE may be harder because: it must characterize and integrate human behavior regarding the interpersonal dynamics, goal-setting, design organization, operations, manmachine interactions, system acceptance and politics. Conclusions Decision Theory and Graph Theory provide two research vectors that strengthen many of Systems Engineering’s present weaknesses, including software integration. Additionally, they are based on previous rigorous work and provide intellectual content to systems engineering, assisting its academic recognition and – hopefully – its intergenerational capture of expertise. Warning: be careful not to overdo the detail and rigor Impacts on Research World usc875 Completed PhD Constraint Theory Decision Theory W. McCumber, UH F. Serrano, MIT R. Pulanna, USC M. Weiler, USC In-Process PhD J. Bartolomei, MIT J. Simpson, UMR T. Henkle, USC Also: Design Sheet at Rockwell Science and Boeing T. Jackson at SJSU on Decision analysis for risk