Optimisation of crop configuration using NSGA-III with categorical genetic operators

One of the main tasks in agriculture is deciding which crop should be planted on which field. Agricultural companies often cultivate dozens of crops on hundreds of fields, making this problem extremely computationally complex. It was solved within evolutionary many-objective optimisation (EMO) framework. Objective functions included: profit, yield risk, price risk, scatteredness, crop rotation and environmental impact (total amounts of fertiliser and pesticide used). As the decision variables were categories (crops) and not real values, NSGA-III was adapted by changing the genetic operators of mutation and crossover from numerical to categorical. Optimisation was performed on the dataset provided by a partnering agricultural company. Out of the resulting population of solutions, characteristic crop configurations were chosen and compared to the benchmark, i.e. company's current strategy.