الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 157 |  |  |
| | | from | 157 | | |
Abstract
Agent technology is a new approach of Distributed ArtificialIntelligence to implement autonomous entities driven by beliefs, goals capabilities, plans, and other agency properties such as adaptationinteraction, and mobility. Multi-Agent System (MAS) have been recentlyused in many simulators and intelligent systems to help humans performseveral tasks. MAS allows the simulation of complex phenomena thatcannot be easily described analyticallyThis work focused on the principles and concepts related to AgentTechnology and Multi-Agent Systems including: the different MachineLearning Algorithms, the effect of using Evolutionary Algorithms (EAto evolve the system and determine the suitable Evolutionary Algorithmthat enhances the system s performance. Also, one of the main objectivesof this thesis is studying the effect of the use of Anticipation in MultiAgent Systems (MAS) that enables the system anticipates undesirablefuture situations in order to avoid themSmart Homes have enjoyed increasing popularity in recent yearsResearch in Smart Homes takes part in many research projects accordingto different points of view and application. This thesis is based on usingone of them that is called Power Management in Smart Home (PMSHsimulatorPMSH simulator is based on the concepts of Multi Agent SystemMAS). On analyzing this simulator some disadvantages are defined. Inorder to get over these drawbacks some modifications on this simulatorare suggested to improve its performance. The effects of the suggestedmodifications are presented. It is shown from the experiments that themodifications improved the simulator s performance. Also, thesemodifications allowed evaluating the performance of each agent in thesimulator. Some evolutionary algorithms for cooperative MAS areapplied on the modified simulator. The performance of the simulator isevaluated with each algorithm. The performance of the simulator ischanged with each evolutionary algorithm. The results proved that theOrthogonal Evolution of Teams (OET) algorithm provide betterperformance in average. It is shown that OET evolutionary algorithm givethe promising results rather than Island algorithm and Team algorithmrespectively in average evaluationAbstractiiThe impact of using anticipation behaviour is presented by adding anAnticipation Module (AM) to the simulator. It is shown from the results of theexperiments that the Anticipation Module enhanced the system performance asit can anticipate unwanted states which require too high power and so try toavoid its occurrence. The simulation has been launched with anticipation and
without anticipation, and then performs simulation using three types of
evolutionary algorithms which are: Island, Team, and OET (OrthogonalEvolution Teams) algorithms, the results of the three algorithms were comparedThe results proved that the Orthogonal Evolution of Teams (OET) algorithmwith Distributed Anticipation Module gave the better results in averageFinally, a suggested Evolutionary Algorithm is called Proposed EvolutionaryAlgorithm. It is based on Island Evolutionary Algorithm using an AnticipatoryModule (AM) and applies it on the simulator which provides good results thanother Evolutionary Algorithms.