{"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/model-based-reinforcement-learning-via-meta","title":"Model-Based Reinforcement Learning via Meta-Policy Optimization","arxiv_id":"1809.05214","date":"2018-09-14","proceeding":null,"authors":["Ignasi Clavera","Jonas Rothfuss","John Schulman","Yasuhiro Fujita","Tamim Asfour","Pieter Abbeel"],"abstract":"Model-based reinforcement learning approaches carry the promise of being data\nefficient. However, due to challenges in learning dynamics models that\nsufficiently match the real-world dynamics, they struggle to achieve the same\nasymptotic performance as model-free methods. We propose Model-Based\nMeta-Policy-Optimization (MB-MPO), an approach that foregoes the strong\nreliance on accurate learned dynamics models. Using an ensemble of learned\ndynamic models, MB-MPO meta-learns a policy that can quickly adapt to any model\nin the ensemble with one policy gradient step. This steers the meta-policy\ntowards internalizing consistent dynamics predictions among the ensemble while\nshifting the burden of behaving optimally w.r.t. the model discrepancies\ntowards the adaptation step. Our experiments show that MB-MPO is more robust to\nmodel imperfections than previous model-based approaches. Finally, we\ndemonstrate that our approach is able to match the asymptotic performance of\nmodel-free methods while requiring significantly less experience.","url_abs":"http://arxiv.org/abs/1809.05214v1","url_pdf":"http://arxiv.org/pdf/1809.05214v1.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":"model-based-reinforcement-learning-via-meta","repo_url":"https://github.com/ray-project/ray/tree/master/rllib","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"model-based-reinforcement-learning","task_name":"Model-based 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=1809.05214","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}