Back to search

University of Western Macedonia

Autonomous vehicle control with brain-computer interface

Abstract

dc:description

Brain-computer interfaces (BCIs) allow people to control and interact with machines by decoding their brain signals, usually collected using electroencephalography (EEG). These systems are very useful for assistive technology because they make it possible to control devices like wheelchairs, robots, and drones using only brain signals. This study focused on the applications of BCIs connected with autonomous vehicles, combining theoretical frameworks with extensive real-world scenarios and experiments. The first part of the study included two systematic reviews: one focused on BCI-controlled ground vehicles and the second on drones and other aerial vehicles. The goal of these extensive reviews was to identify current methodologies, trends, and limitations in real-time BCI applications. Based on the insights and findings of the systematic reviews, two BCI-controlled games were developed to evaluate user adaptation and intra-user analysis in BCI applications. The Muse 2 headset was used to acquire raw data, and the OpenViBE software was used to process the brain signals. EEG signals were classified using Linear Discriminant Analysis (LDA) and transmitted through the Lab Streaming Layer (LSL) for real-time game control. LDA was chosen due to its frequent usage and success in similar research. Healthy volunteers participated in the experiments and played the games to evaluate effectiveness and system responsiveness. In the next part of the study, three extensive real-time experiments were conducted to command an electric wheelchair using EEG. In these experiments, both healthy individuals and individuals with brain or motor disabilities participated, which created a diverse dataset with motor imagery (MI) and Electrooculogram (EOG) commands. Several traditional algorithms for signal processing and classification were employed, including Common Spatial Patterns (CSP), Fast Fourier Transform (FFT), LDA, and Support Vector Machines (SVM), all of which were frequently employed in the literature as identified in the early stages of the research. The experiment protocols started with office navigation and progressed to navigation on a predefined route using only brain signals. Three EEG headsets, Muse S, Emotiv Insight, and Emotiv Epoc, were used for these experiments. A key part of the research was the participant preparation protocol, which focused on high-repetition training. This approach aimed to help users improve, feel more confident, and adapt to controlling the BCI systems. In the final part of the study, a robust Convolutional Neural Network (CNN) algorithm was developed to classify left and right MI hand movements. EEG data were processed and spatially filtered, and then they were transformed into 2D images to be used in the CNN training. The proposed model is not personalized since it was trained on data from fourteen healthy subjects, allowing user generalization. The EEG signals were recorded using the Emotiv Epoc headset, which helped the proposed pipeline to be more versatile and suitable for broader search applications. The results validate the proposed algorithm and show that it can lead to significant progress in real-time and generalized control. This research advances the field of BCI-controlled autonomous systems, enhancing their practicality, adaptability, and inclusiveness in real-world assistive scenarios.

Degree

thesis:*
Grantor dc:publisher
University of Western Macedonia
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Glavas, Kosmas
  • Γλαβάς, Κοσμάς

Subjects

dc:subject × 15

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/59980

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Glavas, Kosmas; Γλαβάς, Κοσμάς. Autonomous vehicle control with brain-computer interface. University of Western Macedonia, 2025. http://hdl.handle.net/10442/hedi/59980