Project #3: Collaborative Learning using Fuzzy Logic (CLIFF) Sophia Mitchell, Pre-Junior, Aerospace Engineering ACCEND College of Engineering and Applied.

1

Project #3: Collaborative Learning using Fuzzy Logic (CLIFF)

Sophia Mitchell, Pre-Junior, Aerospace Engineering ACCENDCollege of Engineering and Applied Science, University of

Cincinnati, Cincinnati, OH

Dr. Kelly Cohen, School of Aerospace Systems

An Extension of Fuzzy Collaborative Robotic Pong (FLIP)

Sponsored by The National Science Foundation Grant ID No: DUE-0756921

2

Outline• Goals & Objectives• Introduction– Fuzzy Logic– Literature Review– Scenario

• Methods • Current Progress & Results• Discussion• Future Work• Timeline

3

Mission Control

Overall Objective

Exploring and exploiting the interactions between humans and intelligent robots to create a synergetic team.

4

Research GoalDevelop a robotic coach that learns from its opponent in order to coach its team to a win in the game of PONG.

Human players provide

uncertainty.

Collaborative robots

Robotic Coach

5

Robotic Team A

GOOD PLAYER

Human or Robotic Team B

Robotic Coach

Research Objective

Coach a “bad” robotic FLIP team until they beat the “good” teamat least 75% of the time

6

Research ObjectiveFLIP team plays the

game

Score!

Coach analysis

Coach decides changes

Coach applies changes to

the FLIP team

7

Fuzzy Logic• Allows classification of variables

for more human-like reasoning.• Common terms• Inputs• Rules• Outputs• Membership Function• Fuzzy Inference System (FIS)

8

Fuzzy Decision Making

Bald Not Bald

Percent of hair on head0 25 50 75 100

9

Type 2 Fuzzy Logic

• Brings uncertainty into the membership functions of a fuzzy set

• Linguistic uncertainties can be modeled that were not visible in Type 1 fuzzy sets

• Allows for more noisy measurements to be quantified

10

Gaussian Singleton Interval Type-2 Fuzzy Inference System (Gauss-INST2-FIS)

• Uses a Gaussian primary membership function (μA(x)) • Constant mean (m)• Variable standard deviation (σ, σ1, σ2)

Equation 1: Variable Gaussian Membership Function

11

Literature Review• Shown us several things:– Type -2 Fuzzy logic is being (slowly) still developed– No paper could be found so far that has both the idea of

a coach and type-2 logic.– Learning many helpful tips with type 2 logic – Benchmark problem resulted from one literature review

article• One MATLAB code is published for Type-2 fuzzy logic

systems– Example problems from textbook

• Spotty topics• Not all types and functions were coded

12

METHODS

13

Methods

• Chose environment (MATLAB)• Complete the Benchmark Problem• Use MATLAB development to create T-2 Fuzzy

players• Create the coach• Develop the team with the coach• Test• Refine

• Chose environment (MATLAB)• Complete the Benchmark Problem• Use MATLAB development to create T-2 Fuzzy

players• Create the coach• Develop the team with the coach• Test• Refine

14

Benchmark Problem Methods

• Model the problem• Solve using type-1 fuzzy logic• Create the type-2 fuzzy logic toolbox in

MATLAB• Test the type-2 logic

15

BENCHMARK PROBLEM

16

The Problem• “Genetic learning and performance evaluation of interval type-2 fuzzy

logic controllers” [2]• Filling a drum with water (controls)• Use pump 1 to control water level in tank 2

17

EquationsA = Cross-sectional drum areaH = Liquid levelQ = Volumetric flow rate into the drumα = Discharge coefficients

18

The method

• Use the dynamic equations outlined in the research paper

• Create the Type 2 functions outlined in the paper

• Carefully note changes in result due to changes in m, δ and membership function position.

• Work with the Type 2 functions to replicate results

19

Why?

• Development of Type-2 Fuzzy Logic Software– Needed for work on CLIFF

• Increased familiarity– Known results verify the created software

• Software will be directly translated into research

• Allows added sophistication due to better understanding of the method

20

RESULTS

21

Their Membership Functions - e

22

My Membership Functions - e

23

Their Membership Functions - edot

24

My Membership Functions - edot

25

26

Results

27

DISCUSSION

28

Discussion

• Type two system produces sensible results

• Benchmark problem simulator brings up a good point about type 1 logic– Compare best possible solutions

29

Conclusions

• Both type-1 and type-2 fuzzy logic are very useful in controls applications– Still not convinced if type-2 is better

• Fuzzy logic is a great tool to use for emulating human reasoning

• Creating a type-2 fuzzy logic toolbox is very time consuming

30

Future Work

• Optimizing type-1 and type-2 results in the benchmark problem

• Bringing T-2FIS into FLIP– Change only part of the membership functions to

type-2– Cascading logic using Type-2– Coach will use Type -2

31

Future Work

• Conferences– Undergraduate Research Forum– AIAA Aerospace Sciences Meeting (ASM) 2014

32

Future Plans

• Continue research in aerospace engineering

• Complete my Bachelors and Masters degrees through the ACCEND program at the University of Cincinnati

• Pursue a PhD• NASA - JPL

Go to space.

33

Acknowledgements

UC AY-REU programDr. Kelly Cohen

MOST-Aerospace Labs

34

References[1] Baklouti, Nesrine, Robert John, and Adel Alimi. "Interval Type-2 Fuzzy Logic Control of

Mobile Robots."Journal of Intelligent Learning Systems and Applications. 4.November 2012 (2012): 291-302. Web. 18 Feb. 2013.

[2] Dongrui Wu, Woei Wan Tan, Genetic learning and performance evaluation of interval type-2 fuzzy logic controllers, Engineering Applications of Artificial Intelligence, Volume 19, Issue 8, December 2006, Pages 829-841, ISSN 0952-1976, 10.1016/j.engappai.2005.12.011. (http://www.sciencedirect.com/science/article/pii/S0952197606000388)

[3] Mendel, Jerry. Uncertain Rule-Based Fuzzy Logic Systems: Introduction and New Directions. Upper Saddle River, NJ: Prentice Hall PTR, 2001. Print.

[4] Castillo, Oscar, and Patricia Melin. Type-2 Fuzzy Logic: Theory and Applications. 1. Heidelberg: Springer, 2008. Print.

[5] Castillo, Oscar. Type-2 Fuzzy Logic in Intelligent Control Applications. 1. Heidelberg: Springer, 2012. eBook.