Year of defence: 2022
Abstract
Humanoids robots are not yet able to safely work in similar environment than humans and to perform the same tasks. The objective of this thesis is to study whole-body control algorithms for humanoids robots in order to perform aircraft manufacturing operations such as drilling. This thesis is part of the joint-lab between the LAAS-CNRS laboratory and the Airbus Operations SAS company. The presented research and innovative studies rely on recent advances on control theory, optimization and stability analysis.The main contributions of this thesis are the design and implementation of new real-time controllers for humanoid robot, integrated on the robot TALOS. The illustrated applications are locomotion and force task operations. First, the actuation chains parameters of the robot are identified in order to directly control the robot in current while having low level protection mechanisms. The proposed solution confirms the capabilities of the robot TALOS to achieve torque control in complex scenario with high payload.Thus, real-time whole-body torque controllers are then investigated. Three controllers are implemented and benchmarked, two in position and one in torque. Their comparison is performed on complex locomotion scenarios using several metrics including trajectory tracking and energetic criteria. The whole-body torque controller is validated in simulations and its advantages compared to the position schemes confirms the decision to implement it for manufacturing operations.However, the first tests realized on the real robot led to a dangerous divergence of the solution. Thus, a stability analysis was performed to ensure the safety and robustness of the solution for industrial applications. The developed solution is based on passivity theory, augmenting the whole-body torque control scheme with a global energy tank monitoring the power flow of the system.This new passive scheme is validated in simulations on locomotion and multi-contact scenarios. The latter involves force applications as a first step toward manufacturing operations.The results of this thesis have been integrated in the Stack-of-Tasks framework of the LAAS-CNRS.