Hybrid U-Net and Fuzzy Logic Framework for Reliable Navigation on Unmarked Roads

Agus Trisanto, Putri Agniya, Arjon Turnip, Dessy Novita, Muhammad Rasyid Ramdhani

Abstract


ABSTRACT  Autonomous vehicle navigation on roads without lane markings presents a significant challenge, as conventional systems relying solely on GPS often fail to accurately determine the vehicle’s lateral position. This study proposes a semantic segmentation approach based on a Convolutional Neural Network (CNN), specifically the U-Net architecture, to recognize road elements using a single camera as the visual sensor. The objective is to develop a real-time navigation system capable of identifying lane positions by classifying images into road and road-edge categories. The segmentation results are then processed using a fuzzy logic controller to determine appropriate steering decisions.
Experimental evaluation shows that U-Net with a MobileNetV2 encoder achieves the highest Mean Intersection over Union (MIoU) of 96.9%. The navigation system performed effectively, achieving a 100% non-fallback navigation rate in straight-road conditions without shadows, while performance decreased to 48% in branching road scenarios wider than 4.5 meters. Speed testing further indicates that the system maintains consistent computational latency, with distance delay increasing linearly with vehicle speed, suggesting an optimal operating range of 5–20 km/h. These results demonstrate the feasibility of combining U-Net segmentation and fuzzy logic for autonomous navigation on unmarked roads.


Full Text:

PDF

References


Akmal Arista, A., Paradifta, K. A., Nugraha, Z. and Priambodo, A. S. (2024). Implementasi fuzzy logic pada kendali robot E-Puck wall following. JEECAE: Journal of Electrical, Electronic, Control and Automotive Engineering, 9(1), 16-21.

Al-Jubouri, A. and Jin, Y. (2020). Challenges of GPS-based navigation for autonomous vehicles. Sensors.

Alam, N. and Dempster, A. (2013). Lane-level navigation using low-cost GPS. IEEE Transactions on Intelligent Transportation Systems.

Badrinarayanan, V., Kendall, A. and Cipolla, R. (2017). SegNet. IEEE Transactions on Pattern Analysis and Machine Intelligence.

Bagloee, S. M., Tavana, M., Asadi, M. and Oliver, T. (2016). Autonomous vehicles: challenges, opportunities, and future implications for transportation policies. Journal of Modern Transportation.

Chen, Z. et al. (2015). End-to-end learning for steering control. Conference on Neural Information Processing Systems (NeurIPS).

Chen, X. et al. (2019). Deep segmentation for unmarked road boundaries. IEEE Intelligent Vehicles Symposium (IV).

Fagnant, R. and Kockelman, K. (2015). The travel and environmental implications of automated vehicles. Transportation Research Part C: Emerging Technologies.

Febrian, Y., Yahya, R. A., Dzaluli, M. I., Priambodo, A. S. and Eng, M. (2023). Implementasi fuzzy logic dengan sistem visual camera pada robot Jetbot sebagai line follower.

Grigorescu, S. et al. (2020). Survey of deep learning for autonomous driving. IEEE Transactions on Intelligent Transportation Systems.

Guan, Y. et al. (2021). Route optimization for autonomous delivery vehicles. IEEE Access.

Hasirci, S. and Hasirci, A. (2017). Fuzzy steering control for autonomous vehicles. Journal of Intelligent & Robotic Systems.

Jo, K., Kim, J., Kim, D., Jang, C. and Sunwoo, M. (2015). Lane detection using GPS and vision. IEEE Transactions on Intelligent Transportation Systems.

Katrakazas, C. et al. (2015). Real-time motion planning for autonomous on-road driving. Transportation Research Part C: Emerging Technologies.

Kim, H. et al. (2020). Autonomous driving in campus environments. IEEE Access.

Long, J., Shelhamer, E. and Darrell, T. (2015). Fully convolutional networks for semantic segmentation. IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

Moussa, A. et al. (2021). Lane boundary detection using deep learning. Sensors.

Proença, P. F. and Simões, M. (2020). U-Net for road segmentation. IEEE Access.

Qureshi, M. R. J. (2019). Road infrastructure issues in rural areas. Rural Engineering Journal.

Ronneberger, O., Fischer, P. and Brox, T. (2015). U-Net: Convolutional networks for biomedical image segmentation. International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI).

Schoettle, B. and Sivak, M. (2015). Potential impact of autonomous and connected vehicles on emissions and energy consumption. University of Michigan.

Srivastava, S. (2020). Autonomous delivery robots in e-commerce logistics. IEEE Transactions on Automation Science and Engineering.

Sun, C., Zhao, H., Mu, L., Xu, F. and Lu, L. (2023). Image semantic segmentation for autonomous driving based on improved U-Net. CMES-Computer Modeling in Engineering & Sciences, 136(1), 787-801. doi: 10.32604/cmes.2023.025119.

Velaga, N. R. et al. (2012). Challenges in rural road navigation. Transportation Planning and Technology.

Yazici, A. and Unel, M. (2019). Vision-based lane detection without road markings. Robotics and Autonomous Systems.

Zadeh, L. A. (1965). Fuzzy logic. Information and Control.

Zhao, L. and Thorpe, C. (2000). Unmarked road detection using texture. IEEE Intelligent Vehicles Symposium (IV).

Zhao, R. et al. (2022). Real-time fuzzy control for low-speed autonomous navigation. IEEE Access.


Refbacks

  • There are currently no refbacks.


Publisher:

Department of Electrical Engineering
Universitas Padjadjaran

Jl. Ir. Soekarno km.21, Jatinangor, Sumedang, Jawa Barat 45363