{"paper":{"title":"RMA: Rapid Motor Adaptation for Legged Robots","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A two-part algorithm lets quadruped robots adapt motor control to new terrains and dynamics in fractions of a second.","cross_cats":["cs.AI","cs.CV","cs.RO"],"primary_cat":"cs.LG","authors_text":"Ashish Kumar, Deepak Pathak, Jitendra Malik, Zipeng Fu","submitted_at":"2021-07-08T17:59:59Z","abstract_excerpt":"Successful real-world deployment of legged robots would require them to adapt in real-time to unseen scenarios like changing terrains, changing payloads, wear and tear. This paper presents Rapid Motor Adaptation (RMA) algorithm to solve this problem of real-time online adaptation in quadruped robots. RMA consists of two components: a base policy and an adaptation module. The combination of these components enables the robot to adapt to novel situations in fractions of a second. RMA is trained completely in simulation without using any domain knowledge like reference trajectories or predefined "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The combination of these components enables the robot to adapt to novel situations in fractions of a second... RMA shows state-of-the-art performance across diverse real-world as well as simulation experiments.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the distribution of terrains and dynamics encountered in the varied terrain generator during simulation training is sufficiently representative of the real-world test conditions so that the adaptation module generalizes without any real-world fine-tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"RMA lets legged robots adapt to unseen terrains and conditions in under a second by pairing a base policy with a learned adaptation module trained entirely in simulation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A two-part algorithm lets quadruped robots adapt motor control to new terrains and dynamics in fractions of a second.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e413063b9abf298bb7e89c2c2fcc3050a53c8afdf0411c601335283888b4b22b"},"source":{"id":"2107.04034","kind":"arxiv","version":1},"verdict":{"id":"70f7980b-7120-4e09-a341-643f8e3a0957","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T13:27:23.460384Z","strongest_claim":"The combination of these components enables the robot to adapt to novel situations in fractions of a second... RMA shows state-of-the-art performance across diverse real-world as well as simulation experiments.","one_line_summary":"RMA lets legged robots adapt to unseen terrains and conditions in under a second by pairing a base policy with a learned adaptation module trained entirely in simulation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the distribution of terrains and dynamics encountered in the varied terrain generator during simulation training is sufficiently representative of the real-world test conditions so that the adaptation module generalizes without any real-world fine-tuning.","pith_extraction_headline":"A two-part algorithm lets quadruped robots adapt motor control to new terrains and dynamics in fractions of a second."},"references":{"count":64,"sample":[{"doi":"","year":2009,"title":"Surrogate-based aerodynamic design optimization: Use of surrogates in aerodynamic design optimization","work_id":"3b5ef2e0-1c7b-44df-b0be-0da2adfc9ab6","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2014,"title":"Rapidly exponentially stabilizing control lyapunov functions and hybrid zero dynamics","work_id":"658c8d7e-e191-43cd-b77d-be50fd31fb1a","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2018,"title":"Fast online trajectory optimization for the bipedal robot cassie","work_id":"d0323a19-450a-463e-9645-c85b0a687dd4","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2018,"title":"Design and Control of Small Legged Robots","work_id":"c096b1a6-5442-4aab-a881-f663ce3a743d","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2018,"title":"Mit chee- tah 3: Design and control of a robust, dynamic quadruped robot","work_id":"ac9a50c6-6691-404f-a234-749856eebb8a","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":64,"snapshot_sha256":"6baf583b92f9a6281781fb678d6c48f139a2acdc1eee9a7ce0691456c60274e2","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"5650028a1d0885e6279b58b42a293e219cb10fe57170a5af4a880196d2d51f6f"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}