• Tejal Deshpande
• Joel Kinny
• Tanmay Malekar

Lane detection, overall trajectory estimation and reliable navigation in an unknown environment are the key challenges of autonomous driving. This paper compares two machine learning based approaches; SVM and MLP, for lane trajectory detection. Key focus is on estimating the scope and applicability of an offline trajectory learning model for real time steering control. The actual indoor navigation is carried out to compare the performance of SVM and MLP algorithms and verify the reliability of navigation for unknown trajectories. Single Raspberry Pi camera has been used for trajectory sensing and path planning. Several real indoor tests are implemented to assess the performance. Results indicate that SVM provides way better accuracy estimation of trajectory curving, therefore the steering angle as compared to an MLP model. Major contribution of this experimentation is the reliability and the study of images captured from low cost Pi-camera and confirming the scope of using it for in-house robotic navigation.

Autonomous navigation, machine learning, image classification, neural networks, camera, robotics

"Comparison of MLP and SVM for Autonomous Vehicle Navigation", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 3, page no.253-263, March-2019, Available :http://www.jetir.org/papers/JETIR1903I42.pdf

"Comparison of MLP and SVM for Autonomous Vehicle Navigation", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 3, page no. pp253-263, March-2019, Available at : http://www.jetir.org/papers/JETIR1903I42.pdf