Parallelization and Grid Integration of a Mobile Robotics Application

Current trends in mobile robotics involve many small yet surprisingly computing-intensive tasks combined in one complex application. While the use of just one personal computer is insufficient in terms of computing power and reliability, the small dimensionality of each of the application’s tasks does not justify supercomputers use. In addition, as more and more robotics laboratories across the world meet the urging need of sharing the robotic resources between them, remote computing becomes increasingly important, with problems such as communication slowdown arising. This work addresses these challenges in the context of a simple mobile robotics application, with plan-based navigation and landmarks based self-localization. • We present our solution for parallelizing and migrating a module of this application on local and remote computers, based on target machine’s architecture. • We present and discuss comprehensive experimental tests. • We discuss the possibility of Grid integration for mobile robotics applications. The main achievements are the use of parallel approach for a module of our application, the proof that common-off-the-shelf computers of today can be used to speedup mobile robotic applications and the promising theoretical aspects of Grid integration of such an application.