Recent advancements in autonomous driving have significantly progressed, particularly in the area of End-to-End driving. End-to-End driving, a method wherein input images captured to determine vehicle control parameters like steering angles and velocity, has emerged as a key focus within the realm of self-driving vehicle research. This study is dedicated to navigating an autonomous bus, developed by the Robotics Research Lab, RPTU Kaiserslautern-Landau, through the diverse and unstructured environment of the RPTU campus using End-to-End driving principles. The primary objective is to assess the model’s performance in navigating pedestrian-rich environments under varying weather conditions. Given the potential hazards presented by pedestrians, initial model testing is conducted within an Unreal Engine simulator. This simulator facilitates the rapid generation of diverse weather conditions for testing. Both the RPTU Kaiserslautern campus and the autonomous minibus are simulated to accurately represent the real campus environment and vehicle. As pedestrian zones lack the typical urban setting of clearly marked streets. Consequently, this research includes the creation of a specialized dataset to train the model. A multi-output regression model, developed using TensorFlow, is utilized to predict the steering angle of the bus. The evaluation process involves the creation of various testing scenarios to measure different parameters. Although initial results within the training scenarios are promising, performance variations are noted when the scenarios are altered. The model undergoes extensive testing across a range of scenarios, with the resulting trajectories and behavioral responses analyzed and depicted graphically. Subsequent testing phases introduces pedestrians into the simulations to further assess model behavior in pedestrian scenarios. This study contributes to the field by exploring the model’s adaptability to complex environments and identifying areas for future improvement.