Research Interests

A. Autonomous learning systems and Intelligent machines.

My future research interests are centralized around the design and construction of systems that can learn and discover from their environment autonomously. In particular, I am interested in

	Understanding the operation principles of autonomous learning systems, regardless of their physical type (humans, animals, or machines).
	Designing computational models (algorithms, computer representations) to implement the above principles,
	Applying the resultant computational models to solving real-world problems.

A typical example of an intelligent machine is a mobile robot which can move safely in an unknown environment, or a robot manipulator which performs several desired industrial applications such as welding and painting on a specified path, material handling, assembly, etc., in an environment with large geometrical uncertainties. With suitable extension and improvement of the techniques and algorithms which I have so far implemented, my effort will focus on the solution of the following problems:

Navigation of mobile robots in dynamic environments: The goal is to develop an autonomous, intelligent robotic system which is able using sensor information to learn how to move safely in unknown environments and how to behave when unexpected events occur.

Robots co-operation: Tasks which are too complicated or impossible for a single robot could be automated using co-operating robots (mobile or articulated manipulators). The goal is to develop an intelligent mechanism to control the co-operation between multiple robots so that to perform reliable and safely the desired tasks.

The above problems will be faced using considerations from natural evolution and techniques from Artificial Life. This type of approaches are characterized as Evolutionary Robotics because they try to develop autonomous robots through an automatic design process involving artificial evolution. Evolutionary Robotics approaches are based on the notions of genetic algorithms and human-animal behaviour. New behaviours (rules) are learned using past experience and environmental feedback mechanisms (reinforcement learning). The first stages of the problems’ solution will be faced and tested by computer simulation. In the next stages the resultant control systems will be applied on real robots.

B. Theory and applications of Evolutionary Algorithms.

Genetic Algorithms (GAs) are robust optimization tools for complex search problems which can cope with discontinuities, non-linearities and even noisy. However, a GA may sometimes have difficulties to converge in an optimum solution. This is mainly due to some phenomena such as the deceptive problem and the problem of premature convergence. Future work will try to solve these problems. The goal will be the development of some balancing mechanisms between the two main characteristics evolved in a GA: the selective pressure on the individual chromosomes and the degree of the population’s diversity. Research will be mainly focused to the following sections:

	Implement and incorporate into the evolutionary process new advanced forms of genetic operators and techniques such as: diploid (haploid) structures of chromosomes, co-operation between chromosomes, pairs which produce multiple offspring, operators of inversion, dominance, duplication and deletion, etc.
	Implementation of auto-adapting mechanisms of some basic characteristics of a GA during the consequence stages of the evolutionary process. For example, auto-adaptation of the crossover and mutation operators, auto-adaptation of the control parameters with emphasis on the population size and the application rate of the genetic operators.
	Implementation of dynamic representation forms of the genotype, e.g. stochastic representation, genotype with variable length, tree representation, matrix representation.

The application of special forms of evolutionary algorithms such as the genetic programming and evolutionary strategies in the solution of known combinatorial NP-complete problems from the area of Computer Science and Operational Research will be examined. Emphasis will be given on the task scheduling problem. The performance and effectiveness of the produced evolutionary algorithms will be compared to that of traditional solution techniques to the problems.

C. Computational Intelligence in Education and Teaching process

The application of new technologies in Education and teaching process is one of my very recent research directions. My interest focuses on the development of computer games (in co-operation with specialists on Education and Pedagogical) which will help students of elementary schools to easily understand concepts and principles of Biology (in the first phase) especially the principles of Darwin’s theory of natural evolution. The main idea is to develop computer games using techniques from Artificial Life such as the genetic algorithms and neural networks.

Research Interests