A Genetic Algorithm to Optimise Stage Reflectors for Self and Other Reflections

Abstract

One of the fundamental challenges in stage acoustic design is striking a balance between Hearing of Self and Hearing of Other. Optimising two objectives at once is exactly what the Non-dominated Sorting Genetic Algorithm (NSGA-II) was designed for. Two Genetic Algorithm (GA) fitness functions have been developed for Hearing of Self and Hearing of Other to assist in the design of an overhead reflecting canopy above an orchestra platform. These are then used to govern the NSGA-II algorithm. Diffraction effects on the panels that make up the canopy have also been studied. Results indicate diffraction attenuations typically in the range of 3 dB at 500 Hz. Including the diffraction calculations can significantly slow the run times of the GA, depending on the desired accuracy. It is shown, however, that the accuracy of these calculations may be safely reduced during the GA's evolving optimisations, provided that more accurate calculations are performed during the final evaluation of the chosen optimal design. Exchanging the time used for diffraction calculations during the optimisation for higher GA population/generation counts leads to better canopy designs.