|Gutowsky, Stefan: Efficient tele-operation of a robot manipulator by means of a motion capture interface. |
Universität Stuttgart, Fakultät Informatik, Elektrotechnik und Informationstechnik, Bachelorarbeit (2015).
33 Seiten, englisch.
|CR-Klassif.||I.1.2 (Symbolic and Algebraic Manipulation Algorithms)|
One of the core challenges of developing an autonomous robotic system is that of establishing appropriate channels for imparting commands to the robotic system. A successful communication infrastructure is an important component not only for a robotic system at the final deployment state, where it would allow the users to specify tasks, but also during the development stages where the robotic system still has to learn desired behaviours and appropriate responses to the environment's events and other external stimuli. In the framework of Learning by Demonstration, positive (and sometimes negative) examples are provided to the system by means of direct execution. Kinesthetic teaching consists in providing such demonstrations by physically moving the robot's body -- alas, it is known to be a cumbersome and time-consuming procedure. Another approach is that of controlling the robotic system more indirectly through tele-operation. We would like to build the appropriate infrastructure to control our PR2 robot using tele-operation, such that a human executer can perform a variety of tasks with minimal effort. We aim at achieving this using our high-precision Polhemus G4 motiontracking system to keep track of human poses and gestures, and interpret these as appropriate commands for the PR2. The ability of switching between various modalities of control is highly desired (e.g in one modality the user controls the position of the robot in the room; in another he controls the arms to perform manipulation; etc). This project touches upon the subjects of task-space control, collision avoidance and compliance, and gesture recognition.
|PDF (15280394 Bytes)|
|Abteilung(en)||Universität Stuttgart, Institut für Parallele und Verteilte Systeme, Maschinelles Lernen und Robotik|
|Betreuer||Toussaint, Prof. Marc; Baisero, Andrea|
|Eingabedatum||25. September 2018|