Um novo método probabilístico de cinemática inversa baseado na RRT em um espaço de dimensão reduzida

Abstract

The evolution of manipulator robots has increased the complexity of their models and applications, requiring that inverse kinematics methods integrated into their control systems have characteristics such as fast convergence, completeness, low computational cost, and the ability to avoid minimal locations and singularities. The probabilistic search method known as Rapidly Random Tree (RRT) is widely used in the area of motion planning and can also be considered in the context of inverse kinematics. In most RRT literature, a tree is applied in the configuration space, which results in a large space for most manipulators. In this thesis, an inverse kinematics technique based on RRT applied directly to the workspace is proposed, which yields a reduced-dimension search space, by representing the tree nodes as joints and the edges as links of the manipulator. This space change allows an intuitive interpretation of the tree parameters, which facilitates the alteration of the different variables that make up the classic RRT algorithm. Through this intuitiveness, a polarization direction in the probability model was developed towards the regions which contain all possible solutions to the problem of inverse kinematics in order to get a lower convergence time without losing its probabilistic completeness. Such characteristics are evaluated through simulated experiments.