Parallelization of FIVE Method on Multicore Embedded System
Nowadays embedded computers have many processor cores. Because of high computation power and low power consumption, the multicore processor platforms are a good support for mobile robot applications. Many tasks are running in parallel in different cores. In real time environment load balancing and scheduling may be a problem for the operating system. The majority of algorithms in embedded system applications are not parallelized because of resources like memory, processors, communication channels and peripherals. The tasks are blocked while they are waiting for these resources. The quality of parallelization can be measured with the Amdahl formula. In this paper the presented application - fuzzy rule interpolation method also called FIVE (fuzzy interpolation based on vague environment) â?? the tasks are easily divided in parallel tasks. The application itself is also a good support for analyses of multiprocessing and conclude the advantages or disadvantages. This method is lightweight enough to run on embedded systems for mobile robots. The FIVE method allows to describe complicated behavior of a mobile robot in relatively simple way. The parallelized tasks are then implemented on the Parallella microcomputer with 2+16 processor cores. The FIVE method runs on the 16 Epiphany cores which is like a coprocessor. The paper analyses based on the Amdahl law the limitation of parallelization in multiprocessing environment.
parallelization, fuzzy interpolation, embedded system, Amdahl law, real time system