Improving the locomotion of a six-wheeled ground robot on different types of terrain using A* algorithm and neural networks
https://doi.org/10.46973/0201-727X_2025_1_51
Abstract
A convolutional neural network is considered for classifying surface types encountered by mobile robots during navigation tasks. The classification of five surface types encountered in Iraq is analyzed: clay, hills, potholes, asphalt roads, and concrete pavements. The architecture of the convolutional neural network (CNN) consists of three convolution blocks with normalization and ReLU activation layers, a pooling layer, a fully connected classification layer after the CNN. The training with 96.62 % accuracy convinced that it is effective. The ray plots show a sharp decrease in loss and improvement in classification accuracy, and the cross matrix confirms successful recognition of most surface types with no misclassification of hills. CNN allows these robots to adapt faster to such complex terrain by dynamically adjusting navigation paths, which significantly improves reliability and autonomy in realworld operations.
About the Authors
I. M. A. Al-Khafaji
Institute of Information Technologies (MIREA) – Russian Technological University; Mustansiriyah University
Iraq
Iraq
Al-Khafaji Israa M. Abdalameer, Chair of Information Information Systems, Postgraduate Student; Department of Natural Sciences, Assistant
Baghdad
A. V. Panov
Institute of Information Technologies (MIREA) – Russian Technological University
Russian Federation
Russian Federation
Panov Alexander Vladimirovich, Chair of Information Information Systems, Candidate of Engineering Sciences, Associate Professor
References
1. Krizhevsky, A. ImageNet Classification Using Deep Convolutional Neural Networks / A. Krizhevsky, I. Sutskever, G. E. Hinton // Advances in Neural Information Processing Systems. – 2012. – DOI 10.1145/3065386.
2. He, Q. Deep residual learning for image recognition / Q. He, S. Zhang, S. Ren, Q. Sun // IEEE Conference on Computer Vision and Pattern Recognition. – 2016. – DOI 10.1109/CVPR.2016.90.
3. Simonyan, K. Very deep convolutional networks for image recognition at large scales / K. Simonyan, A. Zisserman // arXiv preprint arXiv:1409.1556. – 2014. – URL: https://arxiv.org/abs/1409.1556 (date of access: 15.10.2024).
4. Long, J. Fully convolutional networks for semantic segmentation / J. Long, E. Shelhamer, T. Darrell // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. – 2015. – DOI 10.1109/CVPR.2015.7298965.
5. Deep feature learning for discriminative localization / B. Zhou, A. Khosla, A. Lapedriza [et al.] // IEEE Conference on Computer Vision and Pattern Recognition. – 2016. – DOI 10.1109/CVPR.2016.319.
6. Hart, P. E. Formal Basis for Heuristic Determination of Minimum Cost Paths / P. E. Hart, N. J. Nilsson, B. Raphael // IEEE Transactions on Systems Science and Cybernetics. – 1968. – DOI 10.1109/TSSC.1968.300136.
7. Stentz, A. Optimal and efficient path planning for partially-known environments // Proceedings of the IEEE International Conference on Robotics and Automation. – 1994. – DOI 10.1109/ROBOT.1994.351061.
8. Thrun, S. Probabilistic Robotics / S. Thrun, W. Burgard, D. Fox // Cambridge, Massachusetts : MIT Press. – 2005. – 647 p. – ISBN 978-0-262-20162-9.
9. Invitation to 3D Vision : From Images to Geometric Models / Y. Ma, S. Soatto, J. Kosechka, S. S. Sastry // New York. – USA : Springer. – 2004. – 528 p. – ISBN 978-0-387-00893-6.
10. Wong, J. Y. Theory of ground vehicles. 3rd ed. – New York, NY : Wiley, 1989. – 528 p. – ISBN 978-0-471-35461-3.
Review
For citations:
Al-Khafaji I.M., Panov A.V. Improving the locomotion of a six-wheeled ground robot on different types of terrain using A* algorithm and neural networks. Vestnik Rostovskogo Gosudarstvennogo Universiteta Putej Soobshcheniya. 2025;(1):51-57. (In Russ.) https://doi.org/10.46973/0201-727X_2025_1_51
Views:
12
JATS XML
Email this article 