Navegação de um carro autônomo em ambientes urbanos complexos

Abstract

Vehicles were once considered the domain of mechanical engineers. However, unprecedented advances in automobiles and information technology have transformed the traditional vehicle into a large-scale, intelligent, and entertainment-rich computing and mobility machine. Upon closer examination of recent technological advancements, it is discovered that the features and characteristics offered by cutting-edge communication and computing technologies, coupled with the emergence of next-generation cars, provide the foundation for the realization of smart vehicles. These smart cars are autonomous, supporting features such as detecting the surrounding environment, making quick and timely decisions, navigating without human intervention on the road, maintaining safe mobility standards, performing all types of maneuvers, and cruise control, for example. The proposal of this dissertation is to adapt the pure pursuit controller for lateral control and implement a hybrid controller for longitudinal control, combining the two-position controller with predictive control. In this context, when there are no other vehicles on the route, the system prioritizes maximizing speed to optimize travel time. On the other hand, when other vehicles are present, the system’s focus is on ensuring safety and promoting mutual respect among drivers. To evaluate the performance of this driving mode in different scenarios, this work utilizes the CARLA simulator. The results confirm the effectiveness of the controller and serve as a starting point for broader studies and experimental evaluation.