Soft Computing Methods

A tantárgy neve magyarul / Name of the subject in Hungarian: Lágy számítási módszerek

Last updated: 2013. július 1.

Budapest University of Technology and Economics
Faculty of Electrical Engineering and Informatics

MSC programme (Informatics)

Major: Autonomous control systems and robots

Course ID Semester Assessment Credit Tantárgyfélév
VIIIM129 1 2/1/0/v 4  
3. Course coordinator and department Dr. Harmati István,
4. Instructors

Dr. István Harmati

Dr. Bálint Kiss 

5. Required knowledge Mathematics, Control Engineering
6. Pre-requisites
Kötelező:
NEM ( TárgyEredmény( "BMEVIIIMA09" , "jegy" , _ ) >= 2
VAGY
TárgyEredmény("BMEVIIIMA09", "FELVETEL", AktualisFelev()) > 0 )

A fenti forma a Neptun sajátja, ezen technikai okokból nem változtattunk.

A kötelező előtanulmányi rend az adott szak honlapján és képzési programjában található.

Ajánlott:
-
7. Objectives, learning outcomes and obtained knowledge

The goal of the course is to introduce the state-of-the-art soft computing and artificial intelligence methods used in system modeling and control theory. The methods are introduced in the frame of nonlinear identification and control problems.

 

Students successfully satisfying the course requirements are prepared in system modeling and to design and implement control algorithms for complex systems. In general, they are able to contribute to the solution system optimization and decision making problems. They obtain skills to apply fuzzy systems, neural networks, genetic algorithms and swarm intelligence on technological and nontechnological areas (e.g. biology, economics). Also, they are able to take part in the development and research of information system with high demand on artificial intelligence techniques.

8. Synopsis

14 weeks of classes: 26 hours of lectures + 13 hours of classroom practices. Classroom practices illustrate the methods with application examples. The topics of the lectures are the following:

 

  1. Fundamentals of fuzzy-neural systems. Fuzzy implication, defuzzification, Sugeno-type fuzzy systems.

 

  1. The block diagram of Fuzzy Logic Controllers (FLC), the functionality of the blocks, Fuzzy PID and PD controllers.  MacVicar-Whelan meta rules. The rule base design of Fuzzy PD controller. 

 

  1. Overview of numerical optimization methods. The necessary analytical condition of the optimal solution considering the constraints. The statement of optimization problem, aczive set, LICQ condition, the Lagrange function of the optimization problem. First order (Karush-Kuhn-Tucker) conditions.

 

  1. Optimization methods. Gradient-like, conjugate gradient, quasi Newton methods. Computation of gradient in neural networks. Subtractive clustering, computation of gradient in adaptive networks, ANFIS.

 

  1. The architecture of Genetic Algorithms. Linear and nonlinear fitness function, selection, binary and real genetic operators, reinsertation strategies. Multipopulation algorithm. Controller design with genetic algorithm.

 

  1. Adptive fuzzy control. Nominal and supervisory control. Indirect (model based) and direct adaptive control. Stability analysis.

 

  1. Direct adaptive neural control with full state feedback, adaptive control with neural network based nonlinear observer. Case study: flight control.

 

  1. Fuzzy approximation based on SVD. The algorithm, methods to satisfy the mathematical conditions, multivariable extension. Control  design with SVD technique.

 

  1. Optimization and control design with evolutionary programming and bacterial algorithms. The algorithms, guzzy interpretation, control design.

 

  1. Swarm intelligence. Motiovation, common properties, The definition of swarms and intelliogence. Ant Colony Optimization (ACO). The base of global behavior, the mathematical model of ants. The difference between the real and artificial ants. The role of pheromone. The methaheuristic of ACO.

 

  1. Particle swarm optimization (PS). Motiovations, benefits and drawbacks, the concept of PSO, the artificial swarm, optimization algorithm, the motion of particles. Implementation issues, discrete implementation, variants.

 

  1.  Learning algorithms. Algorithms that learn equilibria, best response. Bounfaries in computations, Wolf algorithm and its variants. The control of multiagent systems with learning algorithms.

 

 

  1. Probabilistic model with Bayes networks. 

 

 

9. Method of instruction

26 hours of lectures + 13 hours of classroom practices. 

 

 

 

 


10. Assessment

One midterm is written during the semester, its result must be at least 2 (on the scale of 1 to 5). The result of midterm gives 20 percent in the result of the finale exam. 

 

11. Recaps

The mid-term can be repeated once in the teaching period.


12. Consultations

One week before the midterm if required.


13. References, textbooks and resources

[1] Electronic slides on the educational portal: edu.iit.bme.hu (registration is necessary) 

[2] B. Lantos: Fuzzy systems and genetic algorithms, 2002, Műegyetemi kiadó


14. Required learning hours and assignment
Kontakt óra42
Félévközi készülés órákra15
Felkészülés zárthelyire15
Házi feladat elkészítése
Kijelölt írásos tananyag elsajátítása
Vizsgafelkészülés48
Összesen120
15. Syllabus prepared by

Dr. Béla Lantos, Professor

Dr. István Harmati, Associate Professor

Dr. Gábor Vámos, Associate Professor