UNMANNED GROUND VEHICLE SELECTION WITH ARTIFICIAL NEURAL NETWORKS

Abstract

In recent years, significant advancements have been made in defense systems in response to the increasing demands of countries. The importance of unmanned ground vehicles, a highly critical technology, is becoming more evident with each passing year. In this study, a selection program is intended to be developed to determine the mission purposes for which military unmanned ground vehicles will be used. In line with the operating principles, the basic mechanical systems have been identified. Subsequently, a design catalog containing these basic mechanical systems was created. The desired features for use in the field were asked to the customer. Based on the received responses, the best alternative unmanned ground vehicles were identified using an artificial neural network algorithm. In the artificial neural network model, a feedforward neural network architecture was used. Stochastic Gradient Descent was utilized in the network training function to minimize the model's loss function. The activation functions tanh and softmax were used, and the model has four hidden layers. The model was trained for 150 epochs. Results were obtained for the metrics of accuracy, precision, recall, and F1-score. The model's accuracy rate was found to be %99,63. Such a high accuracy rate indicates that the model has well understood the data in the dataset and provides accurate predictions.