首页|Technological University Researchers Detail New Studies and Findings in the Area of Robotics (PolyDexFrame: Deep Reinforcement Learning-Based Pick-and-Place of Objects in Clutter)
Technological University Researchers Detail New Studies and Findings in the Area of Robotics (PolyDexFrame: Deep Reinforcement Learning-Based Pick-and-Place of Objects in Clutter)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News – New research on robotics is the subjec t of a new report. According to news reporting from Athlone, Ireland, by NewsRx journalists, research stated, “This research study represents a polydexterous de ep reinforcement learning-based pick-and-place framework for industrial clutter scenarios.” Financial supporters for this research include Science Foundation Ireland; Europ ean Regional Development Fund; Higher Education Authority (Hea) on Behalf of The Department of Further And Higher Education, Research, Innovation, And Science ( Dfheris), And The Shared Island Unit At The Department of The Taoiseach. The news correspondents obtained a quote from the research from Technological Un iversity: “In the proposed framework, the agent tends to learn the pick-and-plac e of regularly and irregularly shaped objects in clutter by using the sequential combination of prehensile and non-prehensile robotic manipulations involving di fferent robotic grippers in a completely self-supervised manner. The problem was tackled as a reinforcement learning problem; after the Markov decision process (MDP) was designed, the off-policy model-free Q-learning algorithm was deployed using deep Q-networks as a Q-function approximator. Four distinct robotic manipu lations, i.e., grasp from the prehensile manipulation category and inward slide, outward slide, and suction grip from the non-prehensile manipulation category w ere considered as actions. The Q-function comprised four fully convolutional net works (FCN) corresponding to each action based on memory-efficient DenseNet-121 variants outputting pixel-wise maps of action-values jointly trained via the pix el-wise parametrization technique. Rewards were awarded according to the status of the action performed, and backpropagation was conducted accordingly for the F CN generating the maximum Q-value. The results showed that the agent learned the sequential combination of the polydexterous prehensile and non-prehensile manip ulations, where the non-prehensile manipulations increased the possibility of pr ehensile manipulations.”
NETL
NSTL
