Fusão de imagens e sensores inerciais para a estimação e controle de veículos autônomos

Abstract

The present dissertation proposes a sensoring technique of autonomous vehicles based on the fusion of inertial sensors and data collected from a camera. The autonomous vehicle designed in this project was built using "Mecanum" wheels, which gives the vehicle the capability to move in any direction without having to change orientation. The sensoring system proposed is based on the Extended Kalman Filter using quaternions for the fusion of inertial sensors and computer vision, with the objective of finding the global position and orientation of the system. The inertial measurements used in these systems are made by an accelerometer and a gyroscope. The computer vision aspect of the project is done by a digital camera and an image processing software, which is designed to capture colored points in the image. The theory used to design the vehicle’s controller is based on the Lyapunov’s Stability Theory. This project presents a theoretical basis related to the various elements that compose the system, the mathematical basis used in the filter’s implementation and the controller’s design, a general view of the vehicle’s structure used to validate the theory and the results obtained in practical tests. The system’s performance analysis was based on the analysis of graphics that shows the vehicle’s trajectory, the position and orientation of the system over time and the stability of the proposed control law. The obtained results shows that the proposed objectives were met in a satisfactory manner.