Evolving Artificial Ant Systems to Improve Layouts of Graphical Objects

Artificial ant systems (AAS) have been widely applied to solve many important industrial applications including network configuration or vehicle routing that involve constrained optimisation. In this paper, we explore the possible uses of AAS to handle layouts of graphical objects using constrained optimisation approach. In our evolving AAS, the stepwise movement of each individual ant naturally corresponds to the iterative adjustment of each graphical object in both xand y-coordinates until a local minimum of a predefined objective function is reached. Up to our knowledge, this work represents the first attempt of applying AAS to improve layouts of graphical objects with simpler computation as compared to existing approaches, thus presenting new opportunities for further investigations. 1 Background and Motivation Artificial ant systems (AAS) [1] or their extended ant colony optimisation (ACO) techniques have recently received due attention as a powerful optimisation approach to solve many hard constrained optimisation problems including the network configuration problems [1] and traveling salesman problems (TSPs) [1]. Also widely occurring in many industrial applications, automatic adjustment of graphical objects represents another class of challenging constrained optimisation problems for concise information visualization. Among the many possible alignment problems between the graphical objects in any planar graph, node overlappings are essentially local conflicts that can be effectively handled by local search methods such as a modified EGENET solver [2] proposed to successfully resolve both node and edge overlappings while minimising the predefined objective function on a set of 8 arbitrarily generated test graphs. In this paper, we proposed an interesting AAS Layout search framework using each artificial ant as an autonomous agent to focus on avoiding local conflicts while optimising the objective functions through their indirect communication in the global pheromone matrix. 2 An Empirical Evaluation 0 100 200 300 400 500 0 100 200 300 400 500 (a) The original symmetrical graph 0 200 400 600 80