Recent developments in the field of sensor-based control of robots, together with the availability of low-cost sensor components and of cheap computing power allowed a num- ber of developments in the field of intelligent sensor-based robot devices. These develop- ments are mature enough so that one can envisage to apply them to tele-manipulation tasks. Indeed, tele-manipulation has limitations and it is difficult to perform such tasks as tool loading, tool docking, insertion, screwing and unscrewing, etc.
Within this project we plan to introduce a new mode of tele-manipulatio namely an autonomous mode of operation. Within this new mode, human vision is replaced by computer vision and hence the human being is not any more in the real-time control loop of the manipulator. Instead, the human operator becomes a supervisor at the task level.
The research that we propose to carry out combines computer vision techniques (object recognition, object localisation, tracking, 3-D reconstruction) with robot control techniques in order to be able to perform tasks that require robot fine motions in the presence of uncertainty. Among the innovative aspects of our approach, the ability to operate with an uncalibrated or poorly calibrated camera, and the possibility to deal with large discrepancies between the model of the world and the real world, are the most prominent ones. The project will culminate with a laboratory demonstrator which will be specified in collaboration with an industrial partner.
The candidate was educated in Germany and in USA and the host institute has a great deal of experience in the field of Robotics and in training PhD students.