Path Planning of a Self Driving Vehicle Using Artificial Intelligence Techniques and Machine Vision

Abstract

This paper aims to implement an efficient model of the most optimum path to follow an object on a Self Driving Vehicle (SDV). The path of the vehicle is predicted by using Machine Vision (MV) and Neural networks (NN) model. The NN model uses numerous amounts of training data. First the system works by using the MV algorithms to detect objects with predefined colors. Then, the location of the object is fed to the trained NN to get the speeds of the motors needed to reach the object. The training data are obtained from the manual driving of the vehicle in different experiment settings. In this paper, the neural model is compared with two other methods: object detection using MV model and fuzzy logic (FL) model to prove the efficiency of the neural model. All the three models depend on the live record of the camera board and its fast detection of objects using MV algorithms. The three models showed quite similar results; however, the NN model was much more stable and closer to the optimum path.