[campus icon] Accesskey [ h ] University of Paderborn - Home
DE deutsch
Die Universität der Informationsgesellschaft

GET-Forschungsseminar Abstracts

Guiding Motion Behavior with Attentional Selection-for-Action in a Wheel-Based Robot

Sahil Richard Lopes, GET Lab

Vortrag: 23.05.2018, Raum P 1.4.17


In recent times, much research was conducted in the field of artificial visual attention, which selects the relevant subset of the available visual information for further processing. Up to now, integratio.n of attentional influences such as affordance and saliency with simple robotic actions has only been tested in a virtual reality. The aim of this thesis is to make different attentional influences function as one unit on a physical wheel-based robot platform, for which a host of new challenges must be met. With the attention system and a LIDAR sensor, the robot must be able to guide itself and perform simple actions such as avoiding obstacles or driving towards visually attented targets. In this thesis, the attention influences which had been tested in a simulated environment were tested successfully in real life conditions. These results were compared to those of the simulated environment and were found to be similar. An algorithm was developed which fused the attentional influences such as top-down attention and affordance-based attention that guided the robot towards a particular target while simultaneously avoiding any obstacle that came along the way. This algorithm was tested in a simulated environment and on an actual wheel-based robot. The results were found to be similar in both the environments. In most of the test cases the algorithm was able to guide the robot towards the target. However in some cases it was unable to do so due to the shortcomings of the attentional influences in both simulated and real life environments.