CASPIAN JOURNAL

MANAGEMENT AND HIGH TECHNOLOGIES

Research of models of implementation of the Lee algorithm for a robot motion based on the nvidia architecture in the CUDA technology

Read Kurochkin M. A., Kryshtapovich V. S. Research of models of implementation of the Lee algorithm for a robot motion based on the nvidia architecture in the CUDA technology // Caspian journal : management and high technologies. — 2018. — №3. — pp. 97-108.

Kurochkin M. A. - Cand. Sci. (Engineering), Peter the Great St.Petersburg Polytechnic University (SPbPU), 29 Politekhnicheskaya St., St. Petersburg, 195251, Russian Federation, kurochkin.m@gmail.com

Kryshtapovich V. S. - student, Peter the Great St.Petersburg Polytechnic University (SPbPU), 29 Politekhnicheskaya St., St. Petersburg, 195251, Russian Federation, kry127@yandex.ru

In this paper is described the methods of robot path construction for the NVIDIA GPU architecture by the CUDA technology. There are considered three main memory models of NVIDIA architecture: global, local and texture memory. The algorithm of the robot path construction is based on the Lee algorithm. The algorithm and software implementation executes on the Polytechnic supercomputer (Peter the Great St.Petersburg Polytechnic University). The document resides results of modeling the efficiency of each memory model usage depending on different density of impassable terrain (degree of impassibility) by robot. The density of the impassable terrain in simulation models varies in the range of 5-50%. Paper contains examples of the massively parallel implementation of applications by the CUDA technology. The results will be useful in the applications for planning robots movement over rough terrain with obstacles.

Key words: технология CUDA, распараллеливание алгоритмов, траектория движения робота, волновой алгоритм, модели памяти, CUDA technology, parallel algorithms, robot trajectory, wave algorithm, memory models