Presentation is loading. Please wait.

Presentation is loading. Please wait.

Information-Based Optimization Approaches to Dynamical System Safety Verification Todd W. Neller.

Similar presentations


Presentation on theme: "Information-Based Optimization Approaches to Dynamical System Safety Verification Todd W. Neller."— Presentation transcript:

1 Information-Based Optimization Approaches to Dynamical System Safety Verification Todd W. Neller

2 Focus  Global optimization techniques can be powerfully applied to a class of hybrid system verification problems.  When each function evaluation of an optimization is costly, such information should be used intelligently in the course of optimization.

3 Stepper Motor  Time 

4 Heuristic Search Landscape Make use of simple knowledge of problem domain to provide landscape helpful to search

5 Verification through Optimization  Transform verification problem into an optimization problem with a heuristic measure of relative safety  Apply efficient global optimization

6 Information-Based Approach  Most GO methods waste costly information.  Information-Based Optimization - Previous function evaluations shape probability distribution over possible functions.

7 Multi-Level Local Optimization  Successful methods of comparative study employed two level approach  Generalize to n levels, with each level expediting search for level above  Summarizes information  tractability

8 Comparative Results  Strengths: coarsely plateaued f’, no startup cost for simple functions  Weaknesses: Local optimizations to distant minima, weaknesses of LO procedure 

9 Stepper Motor Results  MLLO-IQ first to succeed with all trials  MLLO-IQ better suited to edge minima  MLLO-RIQ better suited to traversing simple f’ CONSTRYURETMIN 


Download ppt "Information-Based Optimization Approaches to Dynamical System Safety Verification Todd W. Neller."

Similar presentations


Ads by Google