Sharif Digital Repository

Evolutionary algorithms have gradually established a strong foothold as powerful search methods and have been widely applied to solve problems in many disciplines. By the way the performance of these algorithms in hierarchical applications is not satisfying. In order to improve the performance and applicability, numerous sophisticated mechanisms have been introduced and integrated into EAs. In this thesis we have tried to implement a multilayer evolutionary algorithm in order to overcome the problem. The practical results show this improvement any various aspects.

 

Clustering is one of the important problems in machine learning, data mining, and pattern recognition fields. When the considered feature space for data representation is not suitable for discrimination of data groups, the data clustering problem may be a difficult problem that cannot be solved properly. In the other words, when the Euclidean distance cannot describe the dissimilarity of data pairs appropriately, the common clustering algorithms may not be helpful and the clusters show arbitrary shapes and spread in such spaces. Although since the late 1990’s several algorithms have been proposed for finding clusters of arbitrary structures, these algorithms cannot yield desirable... 

Image segmentation is a technique which divides an image into significant parts. The accuracy of this technique plays an important role when it applies on medical images. Among various image segmentation methods, clustering methods have been extensively investigated and used. Since it is an unsupervised method, the existence of a small amount of side-information which is extracted from a specific application (in this case, medical image) could improve its accuracy. Using this side-information in clustering methods introduces a new generation of clustering approaches called semi-supervised clustering. This information usually has a format of pair-wise constraints and can be prepared easily... 

Human brain is one of the most wonderful complex machine which is designed for ever. A huge complex network, composed of neurons as tiny biological and chemical processors which are distributed and work together as a super parallel system to do control and vital activities of human body. Today the main secrecies of operation mechanism in individual neurons as fundamental elements of brain are reasonably understood, but network interactions of this wonderful processors and full understanding of information coding in brain seems elusive and remains as a big challenge in many interdisciplinary fields of science, from biology to cognitive science and engineering.
Thus human brain learning... 

Hardware implementation of Spiking Neural Networks (SNN) can bring about a fast and low cost neural network which has more biological support. Memristor nanodevices are most proposed devices for use as synapses to add dynamic learning to SNNs because of their nano-scale dimensions, low power consumption and memory property. One of the most important bottlenecks in the memristor crossbar based SNNs is system-level simulation of learning process due to its huge memristor equations that should be solved for each sample of time. Due to parallel computation capability of our simulation work, we simulate the circuit on single core CPU and then proposed high performance parallel platforms as... 

The main subject of this project is to present a new algorithm for multiple RF sources localization by means of a group of robots. This problem (locating multiple radio sources) could be utilized in many domains, either military (passive radars) or civil (search and rescue) uses. Confronting detrimental effects of real environment on wave’s propagation pattern is the main challenge in this problem. Optimized movement of robots has been chosen as basic strategy to overcome this challenge. Presenting a general solution for this problem requires researches in four topics. The first one is simulation of environment and sensor with main objective of evaluating obtained algorithms’ performance.... 

Design of a spiking neural network suitable for electronic hardware implementation has been a hot research topic for a long time, in recent years implementation of spiking neural network in real time and biological scale has become important for scientists. This is due to the proximity of these networks to human brain, and the efforts led to the creation of numerous biological models and hardware which has been proposed in this context. This research includes two main parts, the first part is effort to implement a greater number of neurons per device and the second part proposes a communication system among devices to achieve biological scale. This work investigates novel neural elements and... 

In this thesis, the goal is movement recording by observer and learning it to Fuzzy Elastic Matching Machine (FEMM). For this purpose, first using the camera (Microsoft LifeCam HD-3000), the information specified on the man, taken during the move. After preprocessing performed on the data, Data on each node of the FEMM the classified and is given to special FEMM for that movement and the FEMM to be trained by adaptive neuro fuzzy inference system. During each iteration of training, Sensitivity of FEMM for training new movement information is reduced because the FEMM updated correctly. In test part, one movement among all movements that training in all FEMMs is selected and is done by... 

Quadrotors are vertical take-off and landing (VTOL) aircrafts with high maneuver capabilities thanks to their four blades. They are widely used for various applications more than any other unmanned aircraft. Taking photos in military zones and industrial complexes is one such application. Due to the highly noisy radio and GPS (Global Positioning System) signals that are a characteristic of these environments, the task of controlling the quadrotors using them is impossible. An alternative approach that is employed in our research is the use of the aerial images from the on-board camera for autonomous navigation. To this end, we first obtain a reference image for the operation area. These... 

Controlling four legged robots is a complex challenge. The previous works in this field can be divided into two main categories, Analytic method and intelligence method. The analytic method is based on extracting the motion equations of the robot, but in most cases these equations are so complicated, and their implementation is difficult. Another method is intelligence method, which is based on learning algorithms. Despite the improvements in walking robots field, still their walking quality is far from human and animal walking. Researchers believe that this advantage is the result of the learning ability in animals and human. Therefore in this paper the human learning ability is used to... 

The Active Learning Method is in fact an adaptive recursive algorithm which embodies a Multi Input and a Single Output (MISO) system as in a fuzzy combination of several Single Input systems and Single Output systems (SISO), and by utilizing a fuzzy technique of Ink Drop Spread tries to explore and extract input to output transfer function behavior in the system of a single-input to a single-output. Although in simulation state, the speed of this model is set at a much higher speed in comparison to other presented models, still is slower than the processing speed of human brain. In regards to hardware implementation, also there remain the fundamental implementation challenges through more... 

Control of powered lower limb prostheses has a locomotion mode- ependent structure which demands a pattern recognizer that can classify the current locomotion mode and also detect transitions between them in an appropriate time. In the way to achieve this goal, this project presents a locomotion mode recognition system to classify daily locomotion modes consist of level- walking, stair climbing, slope walking, standing and sitting using low-cost mechanical sensors. Since these signals have a quasi-periodic nature, using sequential pattern recognition tools, such as Hidden Markov Model(HMM) improves the recognition performance,because they use sequences of information to make a decision. On... 

In this research, a study on Mirror Neuron and Active learning method was done based on the human capability of using their past knowledge in order to understand new systems faster and with more accuracy. Mirror ALM, A new modeling technique based on ALM was proposed that is capable of merging the IDS planes of an old system in order to improve the output of the modeling for a new system. This new technique was tested on a 3D function, state estimation of an inverted pendulum and finally in control procedure of an inverted pendulum. The results of the tests were compared with the classic ALM method to recognize the advantages and disadvantages of the introduced method. The results showed... 

Modular Robots are group of robots which are made of small components called modules. The advantage of these robots to others is their ability to change physical configuration. Using of these robots in configuration changing due to environmental conditions is popular. While, reconfiguration is one of the most important features in modular robot, it’s the most important concern too. Path planning problem is one of the important problems in robotics .So far, most of presented path planning methods and algorithms are based on fixed-structure and they had little attention to path planning and configuration changing, simultaneously. In this thesis, the Genetic Algorithm is used to find path and... 

Recently, emotions have been widely used in systems that may face uncertainty or unknown situations. Various theories have been put forward for explaining emotions and their roles in mind. The model of emotional masks is a newly developed AI1 paradigm based on Minsky’s model of emotional mind in his recent book “The emotion machine”. This model takes a resource management approach toward modeling the mind and views different processes of mind as resources that need to be managed. The present work leads to the development of an intelligent controller (SCELIC2) that is based on ideas from emotional masks, Self-Organizing Controllers and network-based fuzzy systems. On the way to the... 

The first physical realization of the missing fourth fundamental element of electrical circuits, namely memristor, in 2008 by HP labs triggered an immense amount of research on the capabilities of this element in implementing artificial neurons and artificial brain. In this project we will propose several reinforcement learning-based algorithms that are implemented on a specific memristor-based structure, the memristor crossbar structure. Hence we provide a learning paradigm that resembles the human learning paradigm not only because of the the algorithmic core, which is based on learning from sparse and delayed rewards and penalties, but also because of the hardware over which the... 

In recent years, there have been many efforts for distributing a big and complex duty among some agents in order to do it more simply. Large number of these methods are in area called distributed systems, where several different agents start to work together to do a difficult or complex duty which each of them cannot do it lonely. Combination of distributed systems and artificial intelligence is known as distributed artificial intelligence.
One of the theories that has received attention recently, is called Embodiment. According this theory, in cooperation of a set of agents for performing particular task, expertise might not be integrated in a centralized controller, rather it gradually... 

In Energy Management Systems, data analysis is done based on the information gathered from the electricity utility. The closer the information is to its real value, the better the data analysis in EMS becomes. Hence, improving the accuracy of the collected information leads to an improvement in data analysis in EMS. The information collected from the network consists of the measured values of voltage phasor at each bus, the generation or load power at each bus, and the power flow in branches. Since the total number of sensors are usually less than the total network’s parameters, it is not possible to precisely determine the values of Network’s parameters, and only an estimation of them can... 

Designing artificial intelligence based arithmetic machines that can intelligently perform human-like task has attracted considerable interest among researchers. The ever-increasing advances in soft-computing algorithms require appropriate hardware platforms for such algorithms. One of the most important problems with these algorithms and their hardware implementation structures is the discrepancy between the hardware and the nature of the problem. It can be argued that paying attention to hardware implementation does not necessarily guarantee an optimal implementation of these algorithms. Most of the proposed hardware implementations have very small resemblance to the biological systems... 

The inverted pendulum system has been considered as a well known nonlinear system for testing control algorithms. A two-wheeled balancing robot is a mobile inverted pendulum system whose structure is a combination of a wheeled mobile robot and an inverted pendulum system. Published article studied the Robot in different views. Some papers define it as a vehicle, some other try to model and so many use it to determine the control system. This thesis presents design of fuzzy logic controller for a two-wheeled balancing robot using fuzzy methodes. First we have designed a classic fuzzy logic controller to control both of balancing and Trajectory control of robot. The Fuzzy controller was...