Closed-loop admittance and motion control strategies for safe robotic manipulation tasks subject to contacts

• Autor(a) principal: Mariana de Paula Assis Fonseca
• Data de Publicação: 2021
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Repositório Institucional da UFMG
• Texto Completo: http://hdl.handle.net/1843/37025; https://orcid.org/0000-0003-4083-5355
• topic: Tasks subject to contacts; Interaction forces; Admittance control; Adaptive control; Whole-body control; Dual quaternion; Engenharia elétrica; Quatérnios; Robôs - Sistemas de controle

This thesis contributes to the field of manipulation regarding tasks involving contact with the environment, focusing on safety. To this end, a control architecture is proposed where there is an admittance controller in an outer-loop, which changes the reference trajectory to the robot end-effector to achieve a desired compliant behavior, and a motion controller in an inner-loop used to track this trajectory. More specifically, a six-degree-of-freedom task-space admittance controller using dual quaternion logarithmic mapping is developed in order to impose a desired apparent impedance to the robot. The controller couples the translation and rotation impedance in a single mathematical structure, it is designed based on the energy of the system, and the stiffness matrix is built to be consistent with the task geometry. Furthermore, the formulation is free of topological obstruction, and a solution for the unwinding phenomenon based on a switched error function is presented. Regarding the inner-loop, the choice of the motion controller should take into account the type of actuation of the robot (velocity/position or torque) and the availability of the robot model. On the one hand, if the robot is actuated in torque, appropriate pose controllers for physical interactions are usually based on the robot dynamics as it enables more accurate analyses and helps in the synthesis of the robot dynamic behavior. In those controllers, the ill-conditioning of the joint-space inertia matrix plays an important role. Due to this ill-conditioning, small perturbations in the system can produce large changes in the numerical solutions, which may lead to poor performance or even instability, resulting in unsafe interactions. To overcome these problems, this thesis presents a controller in which the joint-space inertia matrix conditioning is adapted online, consequently enhancing the closed-loop performance. On the other hand, if the robot is actuated in velocity/position, a controller based on the robot kinematics is commonly used. Hence, a kinematic controller based on the dual quaternion logarithmic mapping is also developed. The algorithms proposed in this thesis are validated in simulation and/or experimentally on a robot manipulator and an extension of the proposed architecture for the whole-body case, considering a bimanual mobile manipulator, is also evaluated in simulation. Furthermore, statistical analyses are used to compare the performance of these controllers to other ones of the state of the art, and the results show that the developed techniques are at least as good as or outperforms the ones from the literature.