Navigational control of multiple mobile robots in various environments

Abstract

The thesis addresses the problem of mobile robots navigation in various cluttered environments and proposes methodologies based on a soft computing approach, concerning to three main techniques: Potential Field technique, Genetic Algorithm
technique and Fuzzy Logic technique. The selected techniques along with their hybrid models, based on a mathematical support, solve the three main issues of path planning
of robots such as environment representation, localization and navigation. The motivation of the thesis is based on some cutting edge issues for path planning and navigation capabilities, that retrieve the essential for various situations found in day-to-day life. For this purpose, complete algorithms are developed and analysed for standalone techniques and their hybrid models. In the potential field technique the local minima due to existence of dead cycle problem has been addressed and the possible solution for different situations has been carried out. In
fuzzy logic technique the different controllers have been designed and their performance analysis has been done during their navigational control in various environments. Firstly, the fuzzy controller having all triangular members with five
membership functions have been considered. Subsequently the membership functions are changed from Triangular to other functions, e.g. Trapezoidal, Gaussian functions and combinational form to have a more smooth and optimised control response. It has been found that the fuzzy controller with all Gaussian membership function works
better compared to other chosen membership functions. Similarly the proposed Genetic algorithm is based on the suitable population size and fitness functions for finding out the robot steering angle in various cluttered field.
At the end hybrid approaches e.g. Potential-Fuzzy, otential-Genetic, Fuzzy-Genetic and Potential-Fuzzy-Genetic are considered for navigation of multiple mobile robots. Initially the combination of two techniques has been selected in order to model the controllers and then all the techniques have been hybridized to get a better controller. These hybrid controllers are first designed and analysed for possible solutions for various situations provided by human intelligence. Then computer simulations have been executed extensively for various known and unknown environments. The proposed hybrid algorithms are embedded in the controllers of the real robots and tested in realistic scenarios to demonstrate the effectiveness of the developed controllers.
Finally, the thesis concludes in a chapter describing the comparison of results acquired from various environments, showing that the developed algorithms achieve the main goals proposed by different approaches with a high level of simulations. The main contribution provided in the thesis is the definition and demonstration of the applicability of multiple mobile robots navigations with multiple targets in various environments based on the strategy of path optimisation.