Meta tags:

Headings (most frequently used words):

rl, real, world, approach, via, task, relevant, autonomy, efficient, control, priors, continuously, improving, mobile, manipulation, with, autonomous, our, learns, skills, directly, in, the, without, any, demonstrations, or, simulation, training, abstract, bibtex, learning, practice, overview, auto, grasp, goal, cycles, what, do, we, use, reward, specification, discovery,

Text of the page (most frequently used words):
the (14), with (12), and (10), task (9), #learning (7), #priors (7), real (6), world (6), robot (6), mobile (5), manipulation (5), that (5), for (5), relevant (5), autonomy (5), signal (5), autonomous (4), behavior (4), without (4), efficient (4), data (4), approach (4), from (3), continuously (3), improving (3), object (3), are (3), learn (3), very (3), use (3), control (3), better (3), goal (3), our (3), tasks (3), skills (3), mendonca (2), russell (2), panov (2), emmanuel (2), bucher (2), bernadette (2), wang (2), jiuguang (2), pathak (2), deepak (2), 2024 (2), policy (2), used (2), chair (2), left (2), sweeping (2), right (2), these (2), observations (2), define (2), reward (2), combine (2), note (2), approaches (2), dustpan (2), leveraging (2), collect (2), cycles (2), can (2), stagnation (2), objects (2), grasp (2), supervision (2), learns (2), over (2), practice (2), via (2), human (2), template, nerfies, inproceedings, mendonca2024continuously, title, author, journal, conference, year, bibtex, higher, entropy, sometimes, discovers, new, behaviors, complete, quite, different, prior, discovery, text, prompts, get, masks, segment, anything, moving, combined, depth, obtain, state, estimates, specification, few, samples, while, useful, not, sufficient, reliably, perform, inefficient, all, compared, shown, simple, scripted, standup, movement, restriction, based, distance, detected, planners, rrt, simplified, models, navigation, helps, what, ensure, keep, full, action, space, large, rarely, interacts, leads, leading, isn, yet, performing, standing, auto, how, flexible, ensures, collected, likely, have, overview, show, evolution, practices, learned, hours, present, fully, framework, policies, extensive, instrumentation, this, enabled, which, guides, exploration, towards, interactions, prevents, near, states, basic, knowledge, formulating, generic, rewards, interpretable, semantic, information, low, level, fine, grained, demonstrate, allows, spot, robots, continually, improve, their, performance, set, four, challenging, obtaining, average, success, rate, across, times, improvement, existing, abstract, directly, any, demonstrations, simulation, training, video, paper, carnegie, mellon, university, boston, dynamics, institute, corl,

Text of the page (random words):
continuously improving mobile manipulation with autonomous real world rl continuously improving mobile manipulation with autonomous real world rl russell mendonca 1 emmanuel panov 2 bernadette bucher 2 jiuguang wang 2 deepak pathak 1 1 carnegie mellon university 2 boston dynamics ai institute corl 2024 paper video our approach learns skills directly in the real world via rl without any demonstrations or simulation training abstract we present a fully autonomous real world rl framework for mobile manipulation that can learn policies without extensive instrumentation or human supervision this is enabled by 1 task relevant autonomy which guides exploration towards object interactions and prevents stagnation near goal states 2 efficient policy learning by leveraging basic task knowledge in behavior priors and 3 formulating generic rewards that combine human interpretable semantic information with low level fine grained observations we demonstrate that our approach allows spot robots to continually improve their performance on a set of four challenging mobile manipulation tasks obtaining an average success rate of 80 across tasks a 3 4 times improvement over existing approaches learning via practice we show the evolution of the robot s behavior as it practices and learns skills these are learned over 8 10 hours practice in the real world approach overview task relevant autonomy ensures data collected is likely to have learning signal efficient control use signal to learn skills and collect better data flexible supervision how to define learning signal for tasks task relevant autonomy auto grasp the full action space for the robot left is very large and it rarely interacts with objects our approach leads the robot to grasp objects right leading to data with better signal note that the robot isn t yet performing the task of standing up the dustpan task relevant autonomy goal cycles goal cycles ensure the robot can keep learning without stagnation efficient control what priors do we use leveraging priors helps collect data with better learning signal planners rrt with simplified models for navigation chair movement restriction based on distance to detected object sweeping simple scripted behavior dustpan standup efficient control rl priors we combine rl with behavior priors to learn with very few samples priors while useful are not sufficient to reliably perform the task rl without priors is very inefficient note all compared approaches shown use task relevant autonomy reward specification text prompts used to get object masks from segment anything for chair moving left and sweeping right these are combined with depth observations to obtain state estimates used to define reward rl discovery with higher policy entropy rl sometimes discovers new behaviors that complete the task quite different from the prior bibtex inproceedings mendonca2024continuously title continuously improving mobile manipulation with autonomous real world rl author mendonca russell and panov emmanuel and bucher bernadette and wang jiuguang and pathak deepak journal conference on robot learning year 2024 template from nerfies