Parallel radio-wave propagation modeling with image-based ray tracing techniques

• Published in: Parallel computing 2010-12, Vol.36 (12), p.679-695
• Main Authors: Athanaileas, T.E., Athanasiadou, G.E., Tsoulos, G.V., Kaklamani, D.I.
• Format: Article
• Language: English
• Subjects: Algorithms; Computation; Diffraction; Load balancing; Mathematical models; Message passing interface; Partitioning; Performance evaluation; Radio-wave propagation; Ray tracing; Wireless communication; Wireless communications
• ISSN: 0167-8191; 1872-7336
• DOI: 10.1016/j.parco.2010.08.002

► A parallel ray tracing algorithm for radio-wave propagation prediction. ► Efficient computation decomposition for image-based ray tracing. ► Scalable performance for large databases and for increasing orders of reflection. ► Provide timely detailed analysis of the wireless channel. Ray tracing is a technique based on the numerical simulation of geometrical optics and the uniform theory of diffraction, two well-known approximate methods for estimating a high-frequency electromagnetic field, based on the ray theory of field propagation. Radio-wave propagation prediction models based on ray tracing play an important role in wireless network planning, as they take into account diverse physical phenomena such as reflection, diffraction and foliage attenuation and are considered critical for the analysis of long term evolution (LTE) systems, which requires a detailed description of the wireless channel. A major practical drawback of these models is that they can easily become very computationally expensive, as the required level of accuracy and the corresponding areas of study increase. In this paper, a parallel ray tracing algorithm for radio-wave propagation prediction based on the electromagnetic theory of images is presented. The implementation of the algorithm is based on the message passing interface (MPI). The decomposition of the problem is conducted by partitioning the image tree, while dynamic load balancing techniques are employed by means of the master–worker and the work–pool patterns. The performance of the parallel implementation is studied for different problems and task assignment schemes, showing that high speedups can be achieved.