EML 6229 Introduction to Random Dynamical Systems Mrinal Kumar Assistant Prof., MAE
Syllabus…
Uncertainty: A Fundamental Challenge ✣ Nature is far too complex for engineers ✣ Understanding nature: via math:: analysis via observation:: instruments Both imperfect!! ✣ Outcome: we must live with uncertainty, aka stochasticity Specifically, we must make decisions while limited by stochastic information!
Uncertainty: Examples Weather prediction Hazardous event management Catastrophic event decision making Apophis collision probability in 2029: 2.7% (2004 estimate) Understanding turbulence: affordable transportation 2010 Iceland eruption map
due to practical limitations, e.g. model not good enough, not enough measurements available, neglected effects Characterization of Uncertainty ✣ Given that uncertainty is unavoidable, how do we best capture it in engineering/scientific/financial/economic/sociological systems? Uncertainty Quantification (UQ) Uncertainty Types Epistemic: Aleatory: due to fundamental limitations, e.g. accuracy of instruments computational limits nonrepeatability of experiments Epistemic uncertainty is reducible to aleatory uncertainty in an ideal world
Uncertainty in System Models System (physics)…. to be modeled States … entities that identify/characterize/quant ify the system Accurate math model (no uncertainty) …but too complex!!
Uncertainty in System Models A much simpler, reduced order model, but with uncertainty noise…. Also need: initial conditions: almost always come from measurements
Uncertainty in Measurements ✣ There is always aleatory measurement uncertainty: unavoidable and irreducible ✣ In today’s world, epistemic measurement uncertainty is also dominant (essentially too much information to track, and too few resources) Example: consider the so-called potentially hazardous asteroids
Uncertainty in Measurements ✣ There is always aleatory measurement uncertainty: unavoidable and irreducible ✣ In today’s world, epistemic measurement uncertainty is also dominant (essentially too much information to track, and too few resources) Example: or something closer to home: our space debris View of debris in LEO Expanded view of debris to include HEO plus…. active satellites!! limited resources….
Uncertainty Propagation ✣ When measurements cannot be made (due to lack of allocation), only way to quantify uncertainty is to propagate (forecast) it through use the best known models