Author Topic: Control and System Identification via Swarm and Evolutionary Algorithms  (Read 2821 times)

0 Members and 1 Guest are viewing this topic.

IJSER Content Writer

  • Sr. Member
  • ****
  • Posts: 327
  • Karma: +0/-1
    • View Profile
Author : Tayebeh Mostajabi, Javad Poshtan
International Journal of Scientific & Engineering Research Volume 2, Issue 10, October-2011
ISSN 2229-5518
Download Full Paper : PDF

Abstractó  A central topic of swarm intelligence is the investigation of different types of emergent collective behaviors  in swarms. This article focus on the swarm intelligence applications in control and system identification. Particle swarm optimization (PSO), a novel population based stochastic optimizer with fast convergence speed and simple implementation and genetic algorithm, have been successfully  applied to solve system identification optimization problems. In addition, PSO and ant colony optimization (ACO) have been applied as a navigation algorithm in swarm robots. Some of the recently proposed swarm based metaheuristics such as bacterial foraging optimization algorithm (BFOA), wasp optimization algorithm (WOA), bee optimization algorithm (BOA) and Physarum Solver  will need further investigation to assess their potential for generating state-of-the-art algorithms that are useful for this area.

Index Termsó adaptive control; evolutionary algorithm; global minimum, local minima, robotics; swarm intelligence; system identification.

1   INTRODUCTION                                                                     
SWARM and evolutionary algorithms are useful tools that have been inspired by the natural behavior between organ-isms and their real world interactions or even the laws of physics and the relationship between particles and objects. Maybe the birth of such algorithms began with genetic algo-rithm. GA was presented by John Holland in the 60's AD with the taking idea of behavior of chromosomes in cell divi-sion in living organisms. This is an optimization algorithm that in addition to solving optimization problems, is applied in various applications from music to complex engineering problems [1]. About 1990, another famous algorithm named ant colony optimization (ACO) is introduced by Dorigo Moroco [2] which is inspired from ant behavior where finding the shortest path from the nest to the food source. In 1995 particle swarm optimization (PSO) was born according to behavior of flocks of migratory birds [3]. Since then, many researchers have studied on such algorithms. Some of them found and introduced another novel algorithms. These new algorithms can be divided into three categories: The first was created considering the behavior of other living organisms, for instance, bacterial foraging optimization algorithm (BFOA) mimics how bacteria forage over a landscape of nutrients optimally[4], or bee colony algorithm is inspired  honey bees when they return to the hive and tell the others about finding a good foraging site via the famous dance language, or  principles from self-organized task allocation and social hierarchy within a colony of wasps is modeled as  wasp algorithm for scheduling. or also physarum solver, physarum is a slime mode that is built by a kind of diatom named plasmodium in order to reach to the food optimally [5].  The second category like SOA (seeker optimization

algorithm) is simulated some social human behavior [6]. or Imperialist Competitive Algorithm that is based on dominance of stronger countries on weaker states [7] and the third one has been developed according to fundamental physics laws [8-10].

In fact, all of these algorithms are optimization techniques, some of them like ACO more successful in local and some other like PSO in global optimization problems. Their high ability in solving complex optimization problems Led another group of researchers seeking to apply them directly or as a combination with each other or other computational intelligence algorithms [11] or even conventional methods [12-14,23] to find simpler solutions for solving specialized challenging issues in their own fields. In this regard, it can be pointed to the increasingly influence of swarm and evolutionary computation in control engineering applications.
Nowadays, control engineering has been found many applications in various sections of human life. In control system engineering, the desired output is applied as input to system and the tendency is reaching the desired output and tracking input by output.
A system under the control can be a plain that should track the special path or a robot with special task or even a biological system such as brain, heart or any kind of disabled human body that we want to improve its faults with a suitable controller and even so the system under control can be some sections of social human life such as traffic jam, the fluctuate of burse and other social living difficulties.
Swarm and evolutionary algorithms have been find applica-tions in many applied control engineering. for instance, in various controller designs, path tracking, robotic and swarm robots and also in system identification.
 This article attempts to review some parts of influence of these algorithms in control engineering, controller designing, robotics and specially their applications in adaptive control and system identification.

In order to reach an acceptable control for system or plant, The primary step is, finding a suitable mathematical model. finding a suitable model is the basic work and also the most difficult one in control engineering. If we can obtain a proper model for under the control system, controller designing and good tracking is probable. Vice versa, If we do not have a proper model, we would hardly succeed to design a fine controller and tracking.
In cases where the control system, is a small device and rela-tively detailed map of all its components  are available, an appropriate model can be obtained by using the laws and theories of electricity, magnetism, mechanics, and thermodynamics. But in many cases, such a detailed map is not available. but fortunately there is another approach that is system identification.
In system Identification configuration, in order to find an appropriate model, we need a set of informative data. This data set is produced by applying a proper input to the system and calculate the equivalent output. After that we should do curve fitting and find the best descriptor model for available data set.
Depending on what method to use and what kind of model we selected for identification, several well-known classical methods for estimating unknown parameters of the model have been introduced [15-16]. These methods are very successful in some situations but they do not succeed in many cases efficiently.  Hence, in recent decades, researchers tend to utilize new methods including swarm intelligence and preparing them for system identification.

Read More: Click here...