Controlling a non-holonomic vehicle via artificial neural networks

The use of learning autonomous robots is inevitable in modern production technologies. Modern industries require efficient production, précised measuring and robust control systems due to the hard competition of perfect product manufacturing whereas human wants more comfort in life. Various kinds of robot vehicles for various tasks have been developing increasingly not only in production industry but also in daily life. In this research, a vehicle model with four wheels is built and equipped with actuators, different types of sensors, communication devices, data acquisition and control units in Automatic Control Laboratory of Mechanical Engineering Department of Dokuz Eylul University. On this developed autonomous wheeled mobile robot model, Dempster-Shafer evidence theory is tested at first step in means of sensor fusion for having more reliable data from sensors. Fuzzy logic and most types of artificial neural networks architectures which are popular on autonomous mobile robots are explained starting with basic equations and control parameters, revealing the advantages of ANNs use on autonomous mobile robots in second chapter of this thesis. Chapter three covers the kinematic analysis of mobile robot model. It is the fourth chapter, in which the experimental data and results are demonstrated. The conclusion part, which is named as chapter five, is the part in which the results are evaluated in not only in means of technical or scientific research but also in means of daily life use