Automatic control of an unmanned land vehicle used for disinfection of environments with COVID-19

This study proposes a machine learning-based hybrid method to control an unmanned ground vehicle (UGV) with brain (EEG) signals. In the proposed hybrid method, necessary feature vectors were obtained from EEG signals with wavelet transform and these feature signals were classified using the decision tree model. The signals at the classifier's output are used to move the UGV in the appropriate direction. EEG signal data were taken from 41 people for the training of the system. The data are noise-free with bandpass filters. The performance of continuous and discrete wavelet transform methods has been investigated to obtain the best feature extraction from EEG signals. The training file of the bands with dominant features was created. The training file was trained with the k-Fold Cross Validation method. The trained models were tested on the same test set, and it was tried to predict the right and left direction of the UGV with the data obtained from the EEG signals. It is aimed to find the best performance among the known machine learning methods for right and left direction prediction classification. As a result of the comparison of the data obtained from the models, it was concluded that the decision tree model used with the discrete wavelet transform gave the highest accuracy rate of 83% on the test data compared to other models. In addition, it was concluded that the continuous wavelet transforms with ensemble learning give the highest accuracy rate of 95%.