{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/shared-autonomy-via-deep-reinforcement","title":"Shared Autonomy via Deep Reinforcement Learning","arxiv_id":"1802.01744","date":"2018-02-06","proceeding":null,"authors":["Siddharth Reddy","Anca D. Dragan","Sergey Levine"],"abstract":"In shared autonomy, user input is combined with semi-autonomous control to\nachieve a common goal. The goal is often unknown ex-ante, so prior work enables\nagents to infer the goal from user input and assist with the task. Such methods\ntend to assume some combination of knowledge of the dynamics of the\nenvironment, the user's policy given their goal, and the set of possible goals\nthe user might target, which limits their application to real-world scenarios.\nWe propose a deep reinforcement learning framework for model-free shared\nautonomy that lifts these assumptions. We use human-in-the-loop reinforcement\nlearning with neural network function approximation to learn an end-to-end\nmapping from environmental observation and user input to agent action values,\nwith task reward as the only form of supervision. This approach poses the\nchallenge of following user commands closely enough to provide the user with\nreal-time action feedback and thereby ensure high-quality user input, but also\ndeviating from the user's actions when they are suboptimal. We balance these\ntwo needs by discarding actions whose values fall below some threshold, then\nselecting the remaining action closest to the user's input. Controlled studies\nwith users (n = 12) and synthetic pilots playing a video game, and a pilot\nstudy with users (n = 4) flying a real quadrotor, demonstrate the ability of\nour algorithm to assist users with real-time control tasks in which the agent\ncannot directly access the user's private information through observations, but\nreceives a reward signal and user input that both depend on the user's intent.\nThe agent learns to assist the user without access to this private information,\nimplicitly inferring it from the user's input. This paper is a proof of concept\nthat illustrates the potential for deep reinforcement learning to enable\nflexible and practical assistive systems.","url_abs":"http://arxiv.org/abs/1802.01744v2","url_pdf":"http://arxiv.org/pdf/1802.01744v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"shared-autonomy-via-deep-reinforcement","repo_url":"https://github.com/rddy/deepassist","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"deep-reinforcement-learning","task_name":"Deep Reinforcement Learning"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"},{"task_slug":"reinforcement-learning-2","task_name":"reinforcement-learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1802.01744","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}