|Title||:||Genetic Algorithms and Multi-Objective Optimization in Computer Graphics and 3D Spatial Optimization|
|Speaker||:||Mahesh Chandramouli (Purdue University, USA)|
|Details||:||Fri, 12 Dec, 2014 11:00 AM @ BSB 361|
|Abstract:||:||This discussion delves into the application of multi-objective optimization (MOO) procedures, especially Genetic Algorithms (GA) in computer graphics (CG) and spatial optimization. GAs have been applied with remarkable success in a wide range of applications including industrial design, urban planning, route finding and shortest path, multi-objective site search problems, Traveling Salesman Problem (TSP), automobile design, aircraft design, and process control models. Traditional multi-objective design frameworks are not very efficient in deriving a set of Pareto optimal solutions. When the solution space for design problems is essentially complex and enormous, GA offers a robust mode of balancing computational efficiency, time, and the optimality of solutions. Nonetheless, instead of a single solution, a pool of solutions (all of which are considered equally good) are obtained. Hence, the results in these studies, instead of being merely being presented as a (Pareto) set of candidate solutions, are evaluated using a visualization tool. This helps in the process of informed decision-making, thereby facilitating the selection of the optimum plan by planners, decision-makers, and administrators. Virtual reality (VR) based CG visualization can enable generating multiple points of view (POV), levels of details (LOD) and seeing and understanding hidden information, patterns, outliers, etc.
The second problem being discussed is a computer graphics problem, especially a construction graphics problem involving the interior spatial configuration of built spaces. Systematic configuration of spaces and the arrangement of spatial components in various locations are either absent or minimal either because it is considered trivial (which is incorrect) or due to the computational complexity involved in formulating and solving such multi-objective optimization problems (MOO). In this case, after formulating the genotype and executing the GA for objectives/constraints, the resulting array (Pareto plan) is input to a processing file (.pde- a Java variant), which transforms it to a graphic layout. This iconized graphic interface is used as a reference for creating the 3D Virtual world scenarios. Besides enabling the optimal spatial configuration of the scene elements, this framework also facilitates evaluation and interaction via immersion and navigation in the 3D VR worlds.