AUTONOMOUS VEHICLE GUIDANCE USING NEURAL NETWORK AND RANDOM FOREST MODEL

In this paper, a guidance strategy for autonomous vehicles aimed at preventing collisions caused by faulty sensor data is presented. A kinematic bicycle model is utilized to generate a vehicle motion dataset for training the prediction models. The algorithm employs a random forest prediction model, utilizing a neural network generated dataset to minimize errors in planning the motion. As a case study for implementing the guidance algorithm, a freeway entrance ramp scenario was simulated utilizing MATLAB. The guidance strategy addresses the limitations of GPS-based autonomous vehicle navigation systems, which may experience inaccuracies due to atmospheric conditions and other factors, thereby preventing collisions under such conditions of error.