Improving the genetic algorithm performance in aerial spray deposition management

Determining the parameter value settings to use as input to AGDISP (aerial spray simulation model) in order to produce a desired spray material deposition is considered an instance of the parametric design problem. SAGA (spray advisor using genetic algorithm) was developed to solve this problem. In this paper, we describe several approaches to improve the performance of SAGA. First, we replace the original generational genetic algorithm with a steady-state genetic algorithm, and the original roulette wheel selection with tournament selection. We call the new system SAGA2. Second, we apply a neural network to improve the initial population, crossover, and mutation. We call this version SAGA2NN. Then, we apply GADO, a general-purpose approach to solving the parametric design problem. The integrated GADO version is called SAGADO. Finally, we compare the performance of SAGA, SAGA2, SAGA2NN and SAGADO.