Stochastic Modeling for Mobile Manipulators

Mobile manipulators are valuable and highly desired in many fields, especially in industrial environments. However, determining the end effector position has been challenging for scenarios where the base moves at the same time that the arm follows commands to perform specific tasks. Earlier works have attempted to dynamically evaluate the problem of positioning error for mobile manipulators, but there is still room for further improvement. In this thesis, we devise a dynamical model that leverages stochastic search strategies for mobile manipulators. More specifically, we develop a dynamic model that estimates the position of the robot using an Unscented Kalman filter. Simulations using the Robot Operating System (ROS) and Gazebo were carried out to evaluate our model. Our results for the stochastic method show that it outperforms a deterministic approach (spiral search) under specific Kalman filter covariances of the process and observation noises. Compared to the state of the art, our proposed approach is more robust and efficient, proving to work under different arrangement scenarios with significant better performance.