Vision-based outdoor navigation of self-driving car using lane detection

The evolution of artificial intelligence has served as the catalyst in the field of technology and is making our imaginations real. One of such creation is the birth of self- driving car (autonomous robot). In this paper, a self-driving car physical prototype based on traditional visual method of lane keeping, implemented on Raspberry Pi is proposed which is capable of maneuvering on various kind of tracks autonomously. The proposed method comprises of taking an image from a front facing dashboard camera of the car, detecting the lane from the image and analysing the deviation of the car from the road which is further used to keep the car on the track. The proposed method is implemented on a 1/10 scale car which contains Raspberry Pi 3 Model-B computer and Pi Cam Rev 1.3 for computations and processing. The testing was done without any human intervention on self-made lined track having all kind of turns. The experimental results demonstrate the effectiveness and the robustness of the self-driving car in terms of the navigation error while following the reference trajectory.