Meta tags:
description=A simulator and real robot testbed for Reinforcement Learning;
keywords=KEYWORDS SHOULD BE PLACED HERE;

Headings (most frequently used words):

robot, for, learning, real, reinforcement, teleoperation, data, and, the, domains, sim, air, hockey, manipulation, testbed, with, collection, world, abstract, key, domain, objects, three, 2d, 3d, dataflow, system, modes, human, demonstrations, table, of, evaluations, in, tasks, example, teleoperated, trained, policies, poster,

Text of the page (most frequently used words):
robot (9), and (7), learning (7), for (6), #testbed (6), with (6), tasks (5), real (5), air (5), hockey (5), domains (4), #teleoperation (4), human (4), the (4), reinforcement (4), policies (3), data (3), sim (3), interactive (3), system (2), three (2), challenging (2), based (2), from (2), like (2), two (2), manipulation (2), poster, trained, example, teleoperated, table, evaluations, modes, demonstrations, dataflow, key, domain, objects, promising, tool, complex, even, fast, moving, object, where, hard, coded, might, fail, effectively, reflect, this, category, introduce, dynamic, augmenting, large, family, ranging, easy, reaching, ones, pushing, block, hitting, puck, well, goal, our, allows, varied, assessment, capabilities, also, supports, transfer, simulators, increasing, fidelity, using, dataset, demonstration, gathered, through, systems, virtualized, control, environment, shadowing, assess, behavior, cloning, offline, scratch, abstract, collection, world, arxiv, code, supplementary, paper, novel,

Text of the page (random words):
robot air hockey a novel manipulation testbed for robot learning with reinforcement learning robot air hockey a manipulation testbed for robot learning with reinforcement learning paper supplementary code arxiv teleoperation data collection and real world reinforcement learning abstract reinforcement learning is a promising tool for learning complex policies even in fast moving and object interactive domains where human teleoperation or hard coded policies might fail to effectively reflect this challenging category of tasks we introduce a dynamic interactive rl testbed based on robot air hockey by augmenting air hockey with a large family of tasks ranging from easy tasks like reaching to challenging ones like pushing a block by hitting it with a puck as well as goal based and human interactive tasks our testbed allows a varied assessment of rl capabilities the robot air hockey testbed also supports sim to real transfer with three domains two simulators of increasing fidelity and a real robot system using a dataset of demonstration data gathered through two teleoperation systems a virtualized control environment and human shadowing we assess the testbed with behavior cloning offline rl and rl from scratch the key domain objects the three domains 2d sim 3d sim and real robot dataflow for real robot system teleoperation modes for human demonstrations table of evaluations in domains for tasks example teleoperated data trained policies poster