Development of training tools for haptic teleoperation of a humanoid robot

• Autor(a) principal: Marques, Daniel Filipe da Ponte
• Data de Publicação: 2015
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/16450
• topic: Engenharia mecânica; Visão por computador; Percepção táctil - Interface de computador; Androides

In robotics, the teleoperation of biped humanoids is one of the most exciting topics. It has the possibility to bypass complex dynamic models with learning demonstration algorithms using human interaction. For this procedure, the Humanoid Project at the University of Aveiro - PHUA, ingrained in the production of a 27 degree-of-freedom full body humanoid platform teleoperated by means of haptic devices. The current project also comprises a robot model that has be imported into the Virtual Robot Experimentation Platform: V-REP. The usage of the simulator allows multiple exercises with greater speed and shorted setup times, when compared to the teleoperation of the real robot, besides providing more safety for the platform and the operator during the tests. By using the simulator, the user can perform tests and make achievements towards the reproduction of human movement with the interaction of two haptic devices providing force feedback to the operator. The performed maneuvers have their kinematic and dynamic data stored for later application in learning by demonstration algorithms. However, the production of more complex and detailed movements requires large amounts of motor skill from the operator. Due to the continuous change of users in the PHUA, an adaptation period is required for the newly arrived operators to develop an a nity with the complex control system. This work is focused on developing methodologies to lower the required time for the training process. Thanks to the versatility of customization provided by V-REP, it was possible to implement interfaces which utilized visual and haptic guidance to enhance the learning capabilities of the operator. A dedicate workstation, new formulations and support tools that control the simulation were developed in order to create a more intuitive control over the humanoid platform. Operators were instructed to reproduce complex 3D movements under several training conditions (with visual and haptic feedback, only haptic feedback, only visual feedback, with guidance tools and without guidance). Performance was measured in terms of speed, drift from intended trajectory, response to the drift and amplitude of the movement. Findings of this study indicate that, with the newly implemented mechanisms, operators are able to gain control over the humanoid platform within a relatively short period of training. Operators subjected to the guidance programs present an even shorter period of training needed, exhibiting high performance in the overall system. These facts support the role of haptic guidance in acquiring kinesthetic memory in high DOFs systems.