Characterization of textile grasping experiments
Year: 2012
Type: conference-paper
Abstract: Grasping highly deformable objects, like textiles,\nis an emerging area of research that involves both percep-\ntion and manipulation abilities. As new techniques appear,\nit becomes essential to design strategies to compare them.\nHowever, this is not an easy task, since the large state-space\nof textile objects explodes when coupled with the variability\nof grippers, robotic hands and robot arms performing the\nmanipulation task. This high variability makes it very difficult\nto design experiments to evaluate the performance of a system\nin a repeatable way and compare it to others. We propose\na framework to allow the comparison of different grasping\nmethods for textile objects.\nInstead of measuring each component separately, we there-\nfore propose a methodology to explicitly measure the vision-\nmanipulation correlation by taking into account the throughput\nof the actions. Perceptions of deformable objects should be\ngrouped into different clusters, and the different grasping\nactions available should be tested for each perception type to\nobtain the action-perception success ratio. This characterization\npotentially allows to compare very different systems in terms\nof\nspecialized actions\n,\nperceptions\nor\nwidely useful actions\n, along\nwith the cost of performing each action. We will also show\nthat this categorization is useful in manipulation planning of\ndeformable objects.
Institution: Universitat Politècnica de Catalunya
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W1583822082