Designing limited autonomous robotic systems

Abstract

The aim of this research is to identify and select a cost effective platform and approach to the design and development of low cost autonomous mobile robotic systems. A requirement for such a platform is that it should be of reasonable cost, extendible and that its functionality allows for modification, with reasonable effort. Such an approach may enable organisations with limited resources in funding, limited specialist AI knowledge and scarce hardware expertise, to begin the design of robotic controllers and the application of AI to mobile robots. This solution should prove to be attractive to departments, not historically involved in developing robotic systems, to investigate this area of the application of AI allowing them to apply their own expertise to generate unique solutions to the problem sets. By providing such a platform this research will enable developers to provide more flexible solutions for various application areas for the entertainment, industrial and domestic product markets. The approach was to carry out an in-depth analysis of currently available solutions, modify them as appropriate and evaluate them wherever possible. Testing was carried out by designing robotic controllers for each of these platforms. The platforms investigated include both low cost hardware with embedded software solutions and software simulation environments. The research identified several hardware candidates which featured frequently in publications, these being LEGO’s MINDSTORMS RCX and iRobots Roomba platforms. The use of custom built hardware platforms was not considered for investigation. Eight different simulators were investigated and two main candidates were selected for comparison, “Webots” from Cyberbotics Ltd and “Pyro” an open source simulation project. These two simulators used different approaches and development work and evaluations were carried out on each of them. The thesis arrives at the conclusion that “Pyro” is the most flexible approach of the two environments, but “Webots” is the more established one.