{"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/visual-foresight-model-based-deep","title":"Visual Foresight: Model-Based Deep Reinforcement Learning for Vision-Based Robotic Control","arxiv_id":"1812.00568","date":"2018-12-03","proceeding":null,"authors":["Frederik Ebert","Chelsea Finn","Sudeep Dasari","Annie Xie","Alex Lee","Sergey Levine"],"abstract":"Deep reinforcement learning (RL) algorithms can learn complex robotic skills\nfrom raw sensory inputs, but have yet to achieve the kind of broad\ngeneralization and applicability demonstrated by deep learning methods in\nsupervised domains. We present a deep RL method that is practical for\nreal-world robotics tasks, such as robotic manipulation, and generalizes\neffectively to never-before-seen tasks and objects. In these settings, ground\ntruth reward signals are typically unavailable, and we therefore propose a\nself-supervised model-based approach, where a predictive model learns to\ndirectly predict the future from raw sensory readings, such as camera images.\nAt test time, we explore three distinct goal specification methods: designated\npixels, where a user specifies desired object manipulation tasks by selecting\nparticular pixels in an image and corresponding goal positions, goal images,\nwhere the desired goal state is specified with an image, and image classifiers,\nwhich define spaces of goal states. Our deep predictive models are trained\nusing data collected autonomously and continuously by a robot interacting with\nhundreds of objects, without human supervision. We demonstrate that visual MPC\ncan generalize to never-before-seen objects---both rigid and deformable---and\nsolve a range of user-defined object manipulation tasks using the same model.","url_abs":"http://arxiv.org/abs/1812.00568v1","url_pdf":"http://arxiv.org/pdf/1812.00568v1.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":"visual-foresight-model-based-deep","repo_url":"https://github.com/SudeepDasari/visual_foresight","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"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=1812.00568","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}