Desenvolvimento de um protótipo de veículo autônomo com localização em um ambiente indoor

Abstract

The technological evolution observed in recent years in vehicles and mobile robots has allowed the creation of several automated devices that help people to perform different tasks, such as transporting heavy loads and cleaning rooms. The automation of machines that can move around and act independently also allowed benefits in underwater exploration of deep regions and in interplanetary research and investigation. A critical factor for the navigation of autonomous vehicles is the determination of their location. The knowledge of an estimate of the position and the orientation of the device, allows greater security in the decision making of the performed control. Determine the position of a vehicle is a challenging task, especially when exploring remote places or closed environments, since GPS aid cannot be used on for these cases. From this, this work proposes the development of an autonomous vehicle prototype, which can act manually or autonomously and can navigate as well as determine its position in a closed environment. The developed device can be controlled through a mobile application, where it is possible to send messages through the MQTT protocol, to move and stop the vehicle and also change its operation mode. To estimate the position of the prototype in the environment, a mathematical model was developed for the vehicle movement and this was applied to an Extended Kalman Filter, to use the measurements of distance, acceleration and angular velocity from the sensors to determine the location of the vehicle. Tests in the execution of the different operation modes and in the position determination in different trajectories were carried out, to observe both the functioning of the modes and the position of the vehicle over time, during the movement executing. With the obtained results, it was possible to observe that the identification of the location happened with some errors in the trajectories in certain cases, but that the final position values after the movement were very close to the real values collected. The tests in execution modes detected a satisfactory response from the device in executing the command after sending the message.