Missões autônomas em robôs móveis com tração diferencial: planejamento de caminhos, localização e mapeamento

Abstract

This work presents a design methodology for autonomous missions using mobile robots with differential traction in indoor environments. The missions consist of moving the robot to a goal position starting from an initial pose. For missions success, it is necessary to implement localization and path planning algorithms. For localization, Extended Kalman Filters (EKF) used to fuse odometry with computational vision. The view is responsible for finding artificial markers known as Ar codes that are presented in the environment. The planning is characterized by a hybrid form that corresponds to the union of a deliberative and reactive methods. In the Deliberative Planning, the Direct-DRRT * method, whose base is derived from the Rapidly Exploring Random Tree (RRT), is proposed and used. In addition to the RRT, this planner also presents characteristics of two other methods already presented in the literature, i.e. RRT * and DRRT. The deliberative planner sends to the reactive a set of sub-objectives that connects the initial position to the final goal. Reactive planning is applied during the mission and it is responsible for the dynamic obstacles avoidance that have not been mapped. In the reactive, the Artificial Potential Fields (APF) method is used, which also behaves as the robot controller during navigation. To find the obstacles, we use the sensor Asus Xtion, due to the possibility to find the distances that the locks are in the images generated by this sensor. All the algorithms were implemented through the programming language Python. The results of the work were presented through the simulator Gazebo and practical tests using the P3DX platform. We analyzed the robot behavior in some problems that may occur during navigation, such as kidnapped and local minimum appearance. At the end, the improvement in the results of the Direct-DRRT * path planner is also presented. It is possible to verify the decrease in the time to obtain a path, the number of iterations of nodes and the path length in comparison to other methods. Regarding the location, this dissertation has obtained significant improvements when compared to the methods that use only odometry. Besides, the work was also successful in presenting a solution for the autonomous missions implementation.