VEHICLE CONTROL AND TRAJECTORY PREDICTION USING SUPERVISED NEURAL NETWORK

The research and development in autonomous vehicles have gained massive attention in recent years, intending to transform them into a safe, reliable and intelligent solution for transportation. The latest advancement of Artificial Intelligence and Machine learning techniques finds its application in the development of Autonomous vehicles to make it a resilient system. Perception, Localization, Planning and Control are the four-module that makes up the Autonomous vehicle system as a whole. The final parts of planning and control are the most critical subsystems where the end control decision has to be made. This thesis work is a small contribution towards the research on Supervised Neural Network based trajectory prediction and control models for autonomous vehicles. In this study, the performance of the Multi layer Perceptron Neural Network model for both the trajectory prediction and control is presented. In the first part of the thesis, the construction of the Neural Network Model for the lateral and longitudinal control is presented. In the second part, the architecture of the Neural Network model for the optimal trajectory prediction is briefed along with the MATLAB algorithms utilized for its construction. These two separate Neural Network models are constructed based on the simulation data of the VI-CarRealTime Driver model. Finally, the built-in Neural Network model is tested in the Simulink interface, where the co-simulation with VI-CarRealTime events are performed.