Al-Jarrah, Rami: Development of cooperation between flying robot, ground robot and ground station with fuzzy logic and image processing. 2015
Inhalt
- Abstract
- Zusammenfassung
- Acknowledgments
- Table of Contents
- List of Figures
- List of Tables
- Notations
- Chapter 1 Introduction
- Chapter 2 Literature Reviews
- 2.1. Aerial Robots
- 2.2. Computer Vision on Robot Applications
- 2.3. Tracking and Navigation
- 2.4. Control System
- 2.5. Localize Multi Targets
- 2.6. Conclusion
- Chapter 3 Cooperative Perception, Tracking and Control System
- 3.1. Introduction
- 3.2. Blimp Robot System
- 3.3. Ground Control Station
- 3.4. Fuzzy Sets Model and Possibilities Distributions
- The classical fuzzy semantics are interpretations of fuzzy sets that represent cognitive categories and the system measurements based on linguistic variables. However, the most common and famous problem in the fuzzy control is how to design the member...
- 3.4.1. Introduction to Fuzzy Logic
- 3.4.2. Introduction to Fuzzy Sets
- As it is known, the fuzzy logic is a powerful and strong problem-solving methodology for several applications in control and information processing; more simplicity, it does resemble decision making from approximate data. In addition, it has not only ...
- Definition 11 (α-cut and strong α -cut) [102, 103, 114]: If the fuzzy set A is defined on X and a number α ϵ [0, 1], the α-cut (Aα) and the strong α-cut (Aα+) can be called the crisp sets and they are given by:
- Definition 12 (Support) [102, 103, 114]: If we have set X and A is the fuzzy subset of X, then support of A which is denoted as Supp(A) is the crisp subset of X that has all elements with non-zero membership functions grades in A. The support of A mig...
- Definition 13 (Core) [102, 103, 114]: Lets A be a fuzzy subset of X, the core of A{Core(A)} is the crisp subset of X whose all elements have one membership grades in A. The support of A is the same as the α–cut of A for α =1.
- Definition 14 (Triangular Fuzzy Number) [218]: The fuzzy set could be called triangular fuzzy number if its membership function has the following form:
- Definition 15 (Trapezoidal Fuzzy Number) [102, 103, 114]: The fuzzy set might be called trapezoidal fuzzy, when membership function has following form:
- Where ith is the rules and jth is the input variables. ,a-ij.≤,b-ij.≤,c-ij.≤,d-ij. are known as the breakpoints of the membership functions.
- 3.4.3. Fuzzy Implication Functions
- 3.4.4. Fuzzy Reasoning
- 3.4.5. Fuzzy Rule -Based System
- 3.4.6. A simple example
- 3.4.2. Fuzzy Measures and Evidence Theory:
- The fuzzy measure theory takes into account several special parameters: the plausibility of measures, the belief in these measures, fuzzy membership functions, and classical probability measures. In such theory, even though the conditions are precise,...
- In addition, the evidence theory based on two dual non additive measures: belief measures (Bel) and plausibility measures (Pl). Given a finite universal set Ω and a subset A ∈ P(Ω) then:
- Therefore, these Belief and Plausibility measures can conveniently be defined by a basic probability assignment function m:P,Ω.→[0,1].
- The basic probability assignment is not required that m(Ω)=1 and there is no required relationship between m,A. and m(,A.) [129]. However, the belief and plausibility measures are uniquely determined for all set A∈P(Ω) by the following formulas [129, ...
- where m,B. is the evidence degree or the belief that there is an element belongs to the set A alone, and Bel,A. is the belief that an element belongs to A and also to the various special subsets A. However, the Pl,A. is the belief that an element belo...
- 3.4.3. Possibilities Theory:
- Recently, the fuzzy sets theory has been developed in several directions and applications including the linguistics, logic, pattern recognition, decision-making as well as the system theory. In fact, the common relation between all of these applicatio...
- For simplicity, assume there is a set U, and then the possibility distribution is the characteristic function of a subset E of U. It will model the situation in which all is known about x is that it cannot be exist outside E, thus, (x(u) = 1 if x ϵ E ...
- From a mathematical viewpoint, the information which is modeled by ,𝛑-𝐱.,𝐱. can be nested as random set, and thus; ,𝝅-𝒙.,𝒙. is the function of a random set which might be described as belief function and viewed as random interval I where ,𝛑-𝐱....
- Furthermore, from (3.30 and 3.31) as a special case:
- 3.4.4. Possibility Approach
- In such procedure, the empirical measurement needs a consistency data with semantic aspects of possibility theory and the concept of set statistics requires non-disjoints data. Thus, the frequency data generated from the empirical observations will le...
- Definition 1 [147-150]: In order to derive a possibility distribution from an empirical data, it is indispensable to demonstrate the general measuring record subsets ,𝐔-𝐬.⊆𝛀 as defined below:
- These subsets will be collected for several times and recorded. Where s is counter and i is the number of the element subsets.
- Definition 2 [147-150]: Because the aforementioned data are empirical data and they are unknown, we could not know what and/or how they should be, it is always necessary to eliminate the duplicates from this general measurement record ,𝐔-𝐬. , then d...
- Definition 3 [147-150]: As many empirical data there are a weight value or number of occurrences of ,𝐔-𝐬. in the general measurement record, this is given by the following:
- Definition 4 [147-150]: The set frequency distributions for the data are a function ,𝐦-𝐄.:,𝓕-𝐄.⟶[𝟎,𝟏] which means the number of occurrence for the subset over the total number of the occurrences and it is given by:
- Definition 5 [147-150]: It is known that the mathematical form of the random sets is very complicated. However, in this case of study, we assume them as a finite set, then they will take simply values on subsets of . As a result, the finite random se...
- Where ,𝛇-𝐬. is the evidence function ,ζ-s.=ζ(,U-s.).
- Definition 6 [147-150]: The possibility distribution can be described as possibility histogram which can be obtained as follows:
- Where M is the number of the subsets.
- Definition 7 [147-150]: By assuming each observed subset ,𝐔-𝐬.∈,𝓕-𝐄. is closed interval and symbolized by its statistics left endpoints ,𝐥-𝐬. and right endpoints rs. Thus, the ,𝐔-𝐬.=,,𝐥-𝐬., ,𝐫-𝐬.. and the vectors of endpoints are given by:
- The linear order for the vectors of endpoints ,E. is
- Definition 9 [147-150]: Even though the possibilistic histograms are similar to ordinary stochastic histograms, but they are really performed from overlapping interval observations; thus they are ruled by the mathematics of random sets. If π is a poss...
- Definitions 10 [147-150]: Although the formers of possibilistic histograms are collections of closed segments Tk of different length, but they are not discrete points in stochastic histograms. Thus, instead of the normal interpolation, it is better to...
- For example let us consider that we have done an experiment and collect the observed data. These data categorized into two main intervals A1 = {a, b} and A2= {c, d}. If each intervals observed once that means m(A1)=m(A2)=0.5. The interval ,𝑲-′.=,,,𝒔...
- As it was mentioned previously that the membership functions are trapezoidal and could be designed by possibilities histograms. These kinds of membership functions are general enough and widely used besides they need less computational time compare to...
- Where ,𝑎-𝑖𝑗.,,𝑏-𝑖𝑗.,,𝑐-𝑖𝑗., ,𝑑-𝑖𝑗. are the breakpoints of the trapezoidal membership functions and ,𝐹-𝑖𝑗. and ,𝐻-𝑖𝑗. are the new breakpoints for left and right sides for ith rule and jth input variables. An example for bacterial algo...
- 3.5. Communications
- 3.6. Wireless Sensor Networks Tracking
- 3.6.1. System Model and Problem Formulation
- 3.6.2. Processing at Sensors
- 3.6.3. Processing at the Fusion Center
- 3.6.4. Cramer-Rao Lower Bound (CRLB)
- 3.7 Conclusion
- Chapter 4 Computer Vision for Blimp Robot
- 4.1. Introduction
- 4.2. Image Based Visual Tracking
- 4.2.1. Optical Flow and Appearance Tracking
- 4.2.2. Feature Descriptors and Tracking
- 4.2.3. Robust Matching and Homography
- 4.2.4. Inverse Perspective Mapping
- 4.2.5. Fuzzy Image Processing
- 4.3. Blimp's Embedded Computer Vision
- 4.3.1. Detect and Follow Ground Robot
- 4.3.2. Ground Robot Navigation
- 4.3.3. Blimp Robot Follows 3-D Aerial Object
- 4.4 Conclusions
- Chapter 5 Experiments and Results
- 5.1 Fuzzy Set Model for Blimp's Navigation
- 5.2 Visual Servoing Controller
- 5.2.1 Fuzzy sets model for following Ground Robot.
- 5.2.2 Visual Fuzzy Control for Blimp to Follow 3D Aerial Object
- 5.3 Localize the multi targets
- 5.4 Conclusion
- Chapter 6 Discussion and Future Work
- Appendix A
- OpenEmbedded and Gumstix Boards
- Test the reliability using US MIL-STD-810F Testing
- OpenEmbedded Development Environment
- Appendix B
- Appendix C
- Appendix D
- Appendix E:
- Appendix F:
- Bibliography