Parallel processing with the subsystem synthesis method for efficient vehicle analysis

Abstract

In this paper, parallel processing of a vehicle analysis based on the subsystem synthesis method was developed in a multi-core CPU environment. The subsystem synthesis method provides independent computation of each suspension subsystem module, and so is ideally suited for parallel processing. For parallel program implementation, OpenMP API (Application program interface) for a shared memory multi-core CPU was utilized. Three different parallel implementations were applied to examine the performance of parallel processing. In the first, parallel processing was applied only to the identified parallel regions in the subsystem synthesis method within the integration time loop. In the second, parallel processing was applied from the beginning of the program in order to avoid overhead due to the creation of parallel threads in every time step. The 3rd implementation was the same as the 2nd one, except that unnecessary data synchronization overhead was removed. In order to investigate the performance of the proposed parallel processing, parallel programs with three different implementations were created for a 6 x 6 unmanned robot vehicle model. Rough terrain run simulations were carried out. Performances of the 1st and 2nd implementation methods were worse than that of the sequential program. When parallel processing with proper treatment of overheads was used, a three times increase in computational speed was achieved from the 3rd implementation, compared with serial computation.