Computational & Technology Resources
an online resource for computational,
engineering & technology publications
PROCEEDINGS OF THE FIFTH INTERNATIONAL CONFERENCE ON PARALLEL, DISTRIBUTED, GRID AND CLOUD COMPUTING FOR ENGINEERING
Edited by: P. Iványi, B.H.V. Topping and G. Várady
Single-branch Truss-Z Optimization Based on Image Processing and Evolution Strategy
M. Zawidzki and J. Szklarski
Institute of Fundamental Technological Research of the Polish Academy of Sciences, Warsaw, Poland
M. Zawidzki, J. Szklarski, "Single-branch Truss-Z Optimization Based on Image Processing and Evolution Strategy", in P. Iványi, B.H.V. Topping, G. Várady, (Editors), "Proceedings of the Fifth International Conference on Parallel, Distributed, Grid and Cloud Computing for Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 28, 2017. doi:10.4203/ccp.111.28
Keywords: Extremely Modular System, Truss-Z, discrete optimization, image processing, rasterization, GPU, CUDA, Mathematica, Wolfram Lightweight Grid.
Truss-Z (TZ) is a skeletal system for creating free-form pedestrian ramps and ramp networks among any number of terminals in space. TZ structures are composed of four variations of a single basic unit subjected to affine transformations (mirror reflection, rotation and combination of both). This paper presents a new approach to the optimization of the layout of a singlebranch Truss-Z (STZ) in constrained environment (E). The problem is formulated as follows: create an STZ from a start (sP) to end point (eP) without self-intersections and collisions with two obstacles. This is a multi-criterial optimization problem where three independent objectives are subjected to minimization: the total number of modules (n), the “reaching error” to eP and the “overlapping error”. All three criteria are weighted and aggregated to a single cost function (CF). The calculation of CF is based on image processing of rendered geometry of individual STZs in E. The optimization is performed by population-based classic heuristic method - Evolution Strategy (ES). The computation of CF is the most time consuming, however, its parallelization is rather straightforward. Two parallelization methods are presented: distribution over Wolfram Lightweight Grid and application of general purpose graphical processing units (GPGPUs) with the use of CUDA platform.
purchase the full-text of this paper (price £22)